Article · MarkTechPost. Baseten has released joey-1, a 2B open-weight decision model that returns calibrated probabilities instead of text. It runs at 20 ms P50 latency and costs $0.01 per million input tokens. The post Baseten Releases joey-1, a 2B Decision Model at $0.01 per 1M Tokens appeared first o

Article · llama.cpp. model : Support Limite 1B - Violetto ( #29433 ) gguf-py : add Limite arch Assisted-by: Claude Opus 5.5 Signed-off-by: Giuseppe Scrivano giuseppe@scrivano.org convert : add Limite support Assisted-by: Claude Opus 5.5 Signed-off-by: Giuseppe Scrivano giuseppe@scrivano.org llama : a

Article · llama.cpp. opencl: initial support for iq2 ( #31267 ) opencl: add iq2_s AOS mv opencl: add iq2_s SOA mv and tiled mm opencl: add iq2_s noshuffle gemm/gemv for adreno opencl: add iq2_s dp4a for adreno opencl: optimize iq2_s adreno decode and narrow batches Co-authored-by: Hongqiang Wang wang

Article · noted.lol. Narratarr handles audiobook discovery and downloads while providing a built-in player, all from a self-hosted Docker application.

Article · Immich. chore: version v3.4.0-rc.0

Article · llama.cpp. metal : faster MXFP4 mat-vec, full threadgroups for unary/binary/GLU ( #31075 ) metal : faster MXFP4 mat-vec kernel Decode the fp4 values with the fp16 bit trick instead of a threadgroup lookup table, and load the packed nibbles as aligned 32-bit words. The kernel no longer uses

Article · llama.cpp. hexagon: support for UPSCALE op ( #31331 ) hex-upscale: basic upscale ops (not fully vectorized) hex-upscale: optimize nearest 2x, 4x and bilinear 4x hex-upscale: add support for non-contig rows and dims collapsing hex-upscale: use vgather instead of scalar staging hex-upscale: m

Article · noted.lol. PicPeak is a self-hosted photo gallery platform for sharing events and letting guests contribute their own pictures.

Article · Open WebUI. Added 🧭 Analytics session explorer. Administrators can now open a Sessions tab in Admin Analytics that lists each chat with its tokens, responses, errors and time span, filters them by search, model, token range or sub-agents, and shows a chat's saved replies and tool calls with

Article · n8n. 1.123.90 (2026-10-29) Bug Fixes core: Require a signed token to accept an invitation ( #42292 ) ( 4580f22 ) Performance Improvements Google Drive Trigger Node: Read watched item instead of listing changes ( #42454 ) ( caa8c82 )

Article · n8n. 2.46.2 (2026-10-29) Bug Fixes ai-builder: Open self-healing chat in the result project ( #42466 ) ( 9d781ad ) core: Give each new queue job to one idle worker in turn ( #42400 ) ( d7dbb93 ) core: Include jobs in the claiming list when reading waiting queue jobs ( #42418 ) ( 12257

Article · n8n. 2.46.2 (2026-10-29) Bug Fixes ai-builder: Open self-healing chat in the result project ( #42466 ) ( 9d781ad ) core: Give each new queue job to one idle worker in turn ( #42400 ) ( d7dbb93 ) core: Include jobs in the claiming list when reading waiting queue jobs ( #42418 ) ( 12257

Article · n8n. Co-authored-by: n8n-release-helper[bot]

Article · n8n. 2.45.6 (2026-10-29) Bug Fixes core: Record dynamic time saved for executions in queue mode ( #42482 ) ( f2f0c2f ) editor: Match compact sidebar icon sizes ( #42378 ) ( 56eec49 )

Article · n8n. 2.45.6 (2026-10-29) Bug Fixes core: Record dynamic time saved for executions in queue mode ( #42482 ) ( f2f0c2f ) editor: Match compact sidebar icon sizes ( #42378 ) ( 56eec49 )

Article · llama.cpp. hexagon: improve DDR bw utilization in transposed CONCAT ( #31277 ) hex-concat: double row block size for transposed CONCAT hex-concat: double buffer scratchpads hex-concat: chunked src1 rows in transposed concat hex-concat: pad src1 VTCM stride to spread gather lanes hex-vgather

Article · llama.cpp. HIP/CUDA: enable sparse attn on CDNA Use __ballot for __ballot_sync on old ROCm versions Ignore new large head fattn mma variant

Article · llama.cpp. Requested in the #31125 review. The cases duplicate coverage that is already in the file and add CI time. Test-only, nothing changes in the kernel, the docs or the CMake option. Assisted-by: Claude

Article · llama.cpp. ggml-metal : fix unary kernels compile on older Metal toolchains ( #31 …

Article · llama.cpp. [SYCL] add DPAS flash attention kernel for quantized KV cache ( #31125 ) sycl : add DPAS flash attention kernel for quantized KV cache Decode and small batches (1 to 31 query tokens) now run flash attention straight on q4_0 and q8_0 KV blocks, instead of scalar dequant or convert

Article · llama.cpp. sycl : vectorize reordered-layout dequantization stores (Q4_0, K-quants, IQ3, MXFP4) ( #30976 ) sycl : widen reordered q4_0 dequantization stores The reordered-layout Q4_0 conversion kernel processed one qs byte per work-item with two scalar stores, strided 16 elements apart. Giv

Article · llama.cpp. tests : fix comp_padding CUDA graph reservation ( #31344 ) Assisted-by: Codex Website:

Article · llama.cpp. ggml-openvino: slice view inputs of SCALE ( #31334 ) SCALE now accepts a src0 view with arbitrary row strides. translate_scale used the raw input, and on the dynamic path a VIEW translates to its whole source tensor, so a strided view input scaled the entire source. The model out

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . This release is signed with the key introduced in commit e82a3c6 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way to ens

Article · Ollama. LFM3 on Linux and Windows Liquid AI's LFM3 models, including text-to-speech in lfm3-audio and lfm3-omni , now run on Linux and Windows as well as Apple silicon. Full Changelog : v0.40.12...v0.40.13

Article · llama.cpp. Hexagon: IM2COL improvement - block based HVX parallelization ( #31050 ) hexagon: IM2COL over multiple threads with output-col blocks Assisted-by: Claude Opus 5.5 hex-im2col: fix padding handling hex-im2col: loop unroll common cases Co-authored-by: Max Krasnyansky maxk@qti.qualco

Article · llama.cpp. ggml-cpu: add SVE implementation of ggml_gemv_q8_0_4x8_q8_0 ( #30585 )

Article · Home Assistant. WHAT THE HECK?! November 2026 (function () { var template = document.getElementById("wth-blog-hero"); var image = template && template.parentNode.querySelector(".wth-blog-hero-fallback"); if (!image || !("content" in template)) return; var hero = template.content.cloneNode(true);

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . This release is signed with the key introduced in commit e82a3c6 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way to ens

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . This release is signed with the key introduced in commit e82a3c6 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way to ens

Article · llama.cpp. common, mtmd: bound control vector layer indices and M-RoPE position loops ( #31336 ) common: bound control vector layer indices by the model layer count A direction tensor layer index is parsed from the tensor name with no upper bound, and the buffer size and write offset are co

Article · llama.cpp. opencl: add tiled q4_k GEMM for adreno ( #30412 ) opencl: run the tiled q4_K lm_head through a tiled GEMM at batch > 1 opencl: split tiled dp4a q4_k gemm into its own file and test for dp4a Assisted-by: Codex Co-authored-by: Li He lih@qti.qualcomm.com Website:

Article · Ollama. This temporary patch works around a regression on older versions of MacOS ggml_metal_library_compile_all: failed to build 'unary' library

Article · llama.cpp. hexagon: add conv2d support ( #30995 ) hexagon: add conv2d op hex-conv2d: add double-buffered staging hex-conv2d: full 64-wide activation block transpose straight from staged buffer hex-conv2d: implicit gemm hex-conv2d: remove scalar path and kernel hex-conv2d: unify conv2d kerne

Article · llama.cpp. Speed up IQ4 quant creation with pre-computed halfway points ( #31201 ) Website:

Article · Ollama. llama.cpp: version bump b11963 ( #19360 )

Article · llama.cpp. ggml-cpu: SVE 256 bit implementation of gemm_q6_K_8x8_q8_K kernel ( #27472 ) q6_k GEMM working cleanup review comments review comment addressed - removed pg16_8 incorporated the review comments - replaced switch with if and other points marking q8_k_blocklen UNUSEDto disable warn

Article · MarkTechPost. MiniMax has open-sourced MCode Stack, an MIT-licensed repo of 17 SKILL.md skills and 5 checklists that its MiniMax Code team uses daily for code review, subagent splits, context budgets and long-running commands. The post MiniMax Open-Sources MCode Stack: 15 Agent Skills Its Codi

Article · llama.cpp. ci(s390x): switch to Ubuntu 26.04 runners ( #31317 ) Website:

Article · llama.cpp. mxfp4 sve 256 impl completed working with pretty good perf improvement review comments addressed review comments for SVE 256 check trailing white spaces removed Co-authored-by: Anubhav Srivastava anubhav.srivastava@fujitsu.com

Video · Techno Tim. Video from Techno Tim.

Article · llama.cpp. llama : fix DSV4 compressor padding for simple batches Count actual tokens per sequence when padding compressor writes. Add a graph reuse regression test for odd-sized DeepSeek V4 and V4.1 batches. Assisted-by: Codex tests : cover multiple sequences in DSV4 padding regression Exe

Article · llama.cpp. sycl: work-group argmax for short TOP_K rows ( #31110 ) sycl: work-group argmax for short TOP_K rows Used angelX to optimize SYCL/drama 3 coordination, results are as follows Rows with ncols Arc Pro B70, oneAPI 2026.1, test-backend-ops perf (existing short-row cases), base 9b56cc

Article · llama.cpp. cuda: fuse MoE (MUL_MAT_ID) gate/up and GLU in MMQ ( #31188 ) Website:

Article · llama.cpp. Overview This release adds the LiquidAI LFM3 family (text, vision, audio and omni) together with a new experimental streaming-audio API in mtmd , and wires realtime audio in and out of the server through the new /v1/realtime and /tts endpoints. A batch of new architectures landed

Article · llama.cpp. llama.cpp : bump version to 0.8.0 ( #31309 ) llama.cpp : bump version to 0.8.0 authors : update Website:

Article · noted.lol. Fileship combines ShareX-style screenshot uploading with a self-hosted file library, making uploads easier to manage long after they're shared.

Article · llama.cpp. server : add tool calls to /v1/realtime Assisted-by: Claude Opus 5.5 refactor Co-authored-by: Xuan Son Nguyen son@huggingface.co

Article · llama.cpp. Article from llama.cpp.

Article · llama.cpp. tests : adjust constants to reduce irrelevant warnings ( #31254 ) Website:

Article · llama.cpp. cuda : add top-k MoE fusion for 384 experts ( #31288 ) Assisted-by: Codex Website:

Article · vLLM. Signed-off-by: grYe99 guorongye99@gmail.com Co-authored-by: grYe99 guorongye99@gmail.com (cherry picked from commit bfa2a43 )

Article · llama.cpp. chat : stop qwen3-coder reasoning and content before template separators ( #30557 ) chat : strip trailing whitespace from parsed reasoning and content Assisted-by: Claude Fable 5.1 chat : stop qwen3-coder reasoning and content before template separators Assisted-by: Claude tests

Article · llama.cpp. fix: clamp the A prefetch to end_k in the Vulkan coopmat1 int8 matmul, like the B prefetch fix: zero the A scale for k steps past end_k in the Vulkan coopmat1 int8 matmul instead of relying on the clamped block

Article · n8n. 3.0.1 (2026-10-28) Bug Fixes core: Keep the ExpressionError name in minified builds ( #42309 ) ( b3192c6 ) core: Keep undefined values in QuickJS expression results ( #42264 ) ( 532f645 ) editor: Align carets within split buttons ( #42203 ) ( 61cbfea ) editor: Improve first load

Article · n8n. 1.123.89 (2026-10-28) Bug Fixes Google Drive Node: Fix infinite pagination loop in v1 API request ( #42124 ) ( 3faa762 ) MCP Server Trigger Node: Evict idle sessions to prevent memory leak ( #42120 ) ( 54eaf65 )

Article · n8n. 2.46.1 (2026-10-28) Bug Fixes core: Improve removed-node and Edit Image entries in v3 migration report ( #42270 ) ( d73a148 ) core: Keep the ExpressionError name in minified builds ( #42308 ) ( 4068bb0 ) core: Keep undefined values in QuickJS expression results ( #42263 ) ( 3a6ce

Article · n8n. 2.45.5 (2026-10-28) Bug Fixes core: Keep the ExpressionError name in minified builds ( #42309 ) ( b3192c6 ) core: Keep undefined values in QuickJS expression results ( #42264 ) ( 532f645 ) editor: Align carets within split buttons ( #42203 ) ( 61cbfea ) editor: Improve first load

Article · LiteLLM. Verify this release cosign verify \ --key

Article · ComfyUI. Article from ComfyUI.

Article · llama.cpp. server : add realtime conversation and transcription on /v1/realtime ( #31269 ) server : add realtime conversation and transcription on /v1/realtime Co-authored-by: Xuan Son Nguyen son@huggingface.co Assisted-by: Claude Opus 5.5 examples : default the realtime client to the serve

Article · llama.cpp. model : add LiquidAI/d1-2B and LiquidAI/d1-omni-2B ( #31275 ) model : add LiquidAI/d1-2B and LiquidAI/d1-omni-2B Assisted-by: Claude Opus 5.5 convert : keep the speech task templates out of the d1 GGUFs Assisted-by: Claude Opus 5.5 small refactoring: no_batching Co-authored-by: X

Article · llama.cpp. server: expose the reasoning effort levels of the chat template ( #31271 ) server: expose the reasoning effort levels of the chat template The capability probe renders a user turn for each candidate level (minimal, low, medium, high, xhigh, max) and for an unknown level. A level

Article · llama.cpp. model : add LiquidAI/LFM3-Audio-2B and LiquidAI/LFM3-Omni-2B ( #31257 ) model : add LiquidAI/LFM3-Audio-2B and LiquidAI/LFM3-Omni-2B Co-authored-by: Xuan Son Nguyen son@huggingface.co Assisted-by: Claude Opus 5.5 convert : read the LFM3-Audio config.json as released A transformer

Article · llama.cpp. common : support deferred tools and tool_reference content parts ( #31256 ) common : support deferred tools and tool_reference content parts Assisted-by: Claude Opus 5.5 common : credit the deferred tools co-author Co-authored-by: Aldehir Rojas 765227+aldehir@users.noreply.github

Article · llama.cpp. model : add LiquidAI/LFM3-1.7B and LiquidAI/LFM3-VL-2B ( #31252 ) model : add LiquidAI/LFM3-1.7B and LiquidAI/LFM3-VL-2B Co-authored-by: Sigbjørn Skjæret sigbjorn.skjaeret@huggingface.co Co-authored-by: Xuan-Son Nguyen son@huggingface.co Assisted-by: Claude Opus 5.5 tests : set t

Article · n8n. fix(agents): Show the real tool name for call_tool calls in the stream

Article · n8n. Conflicts: packages/@n8n/instance-ai/docs/architecture.md

Article · n8n. feat(ai-builder): Run SDK source lint in the build and remove the sep…

Article · Ollama. Liquid AI LFM3 LFM3 is Liquid AI's new family of small models built for local agents, tool calling, and structured output, on Liquid's new tri-hybrid architecture. The family includes: lfm3 : a 1.7B text model with a 256K-token context window lfm3-vl : adds image understanding lf

Article · llama.cpp. llama : build the k-pool sparse selection independently of the kq_mask shape ( #31243 ) llama : build the k-pool sparse selection independently of the kq_mask shape qwen4exp (QSA) and glm5-next pick cache cells by index and no longer read the kq_mask values, but they still shaped

Article · LiteLLM. Verify this release cosign verify \ --key

Article · LiteLLM. Verify this release cosign verify \ --key

Article · LiteLLM. Verify this release cosign verify \ --key

Article · LiteLLM. Verify this release cosign verify \ --key

Article · llama.cpp. rpc : catch std::length_error in recv_msg ( #31204 ) Website:

Article · llama.cpp. tests : speed up tensor data generation and model saving ( #31145 ) tests : speed up tensor data generation and model saving Refactor save_models to create only the model, not a llama_context, which is not needed for saving. Replace the slow std::normal_distribution draw with a s

Article · Ollama. mlx: add LFM3 and d1 models ( #19333 )

Article · llama.cpp. vulkan: fix CI llvmpipe by allowing dim 3 broadcast in MUL_MAT ( #31239 ) vulkan: allow dim 3 broadcast in MUL_MAT supports_op rejected every MUL_MAT whose src0 and src1 differ in ne3, although mul_mm, mul_mmq and mul_mat_vec already broadcast src0 over dim 3 through broadcast3.

Article · llama.cpp. cmake: WASI doesn't support spawning ( #30751 )

Article · llama.cpp. mtmd-cli: Reduced scope of batch pointer ( #31155 ) ( #31165 ) mtmd-cli: Reduced scope of batch pointer ( #31155 ) mtmd-cli: Remove CLI regression test Website:

Article · n8n. docs(ai-builder): Point the testmax skill at the A/B benchmark (no-ch…

Article · llama.cpp. server: add router device memory margin parameter for dynamic unloading ( #21231 ) server: add --models-memory-max parameter to allow dynamically unloading models when they exceed a memory size threshold estimate with to-be-loaded model size included use no_alloc to get memory re

Article · llama.cpp. common : --model or -hf is required ( #31242 ) Assisted-by: Codex Website:

Article · llama.cpp. model: Add MiniCPM-OCR-Jev ( #30980 ) model : add MiniCPM-OCR-Jev support The model is a Qwen3.5-VL trunk with a value head read at the last token: it scores an OCR text against an image rendered from it, it does not generate. Add the openbmb-minicpm-ocr-jev decision type, served

Article · llama.cpp. metal : add GGML_OP_TOP_K_MASK ( #31230 ) metal : add GGML_OP_TOP_K_MASK Port the fused top-k mask from the CUDA backend: a grid-stride kernel fills the output with -inf, then one threadgroup per row copies the mask values of the kept cells over it. Function constants drop the sc

Article · llama.cpp. glm5-next : QSA without kq_mask (+ DSA) ( #31009 ) glm5-next : QSA without kq_mask DSA: dsa top-k mask change tested with layers 0,3 of DeepSeek-V3.2 Website:

Article · noted.lol. Keep your homelab documentation in Markdown with Hold My Docs, a self-hosted wiki that saves every change to Git.

Article · n8n. Co-authored-by: Claude Opus 5.5 (1M context) noreply@anthropic.com

Article · llama.cpp. ci : use ccache depend mode for ROCm release builds ( #31231 ) Website:

Article · llama.cpp. ggml : relax scale to non-contiguous views ( #31146 ) ggml : relax scale to non-contiguous views SCALE requires src0 to be 1D-contiguous. Relax this to dim-0 contiguity with arbitrary row strides (ggml_is_contiguous_rows), same as the unary ops in the Metal backend. ggml: relax t

Article · n8n. 3.0.0 (2026-10-27) Bug Fixes Add file-type tooltip to artifact download button ( #41490 ) ( 4a74f02 ) ai-builder: Check execution:read in Assistant execution reads ( #41863 ) ( a967486 ) ai-builder: Send Assistant model calls through the configured AI transport ( #41948 ) ( 9b635

Article · n8n. 2.45.4 (2026-10-26) Bug Fixes core: Recreate removed n8n-sandbox sandboxes and rerun workspace setup ( #41919 ) ( 11e279a ) core: Start sub-workflow error workflow when a pre-run check fails ( #41977 ) ( 9595142 ) editor: Match navigation plus button hover styling ( #41910 ) ( 47

Article · n8n. 2.46.0 (2026-10-27) Bug Fixes Add file-type tooltip to artifact download button ( #41490 ) ( 4a74f02 ) ai-builder: Check execution:read in Assistant execution reads ( #41863 ) ( a967486 ) ai-builder: Send Assistant model calls through the configured AI transport ( #41948 ) ( 9b63

Article · llama.cpp. tests : fix multi-seq split replay and partial-state cases for trained models ( #31216 ) tests : fix multi-seq split replay against trained models The rollback snapshot is captured by the graph that decoded the rolled-back tail, while the reference context decodes the same prefix

Article · llama.cpp. llama : remove obsolete n_pad from memory instances ( #31200 ) Every caller passes n_pad = 1 and the only two users clamp it to 256. Drop the parameter, the members and the trivially true kv_size assert. Assisted-by: dsh:llama.cpp/ggml-org/Qwen3.8-Flash-Next-GGUF (MacBook, M5 Max

Article · llama.cpp. opencl: refine mul mat id -1 index ( #31170 ) Assisted-by: Codex Website:

Article · llama.cpp. opencl: initial support for q3_k ( #31172 ) opencl: add q3_k AOS mv opencl: add q3_k SOA mv and tiled mm opencl: add q3_k noshuffle gemm/gemv for adreno opencl: add q3_k dp4a gemm for adreno Co-authored-by: Hongqiang Wang wangh@qti.qualcomm.com opencl: tune q3_k kernels for adren

Article · ComfyUI. What's Changed model_prefetch: syncrhonize all graph cleanups by @rattus128 in #17178 Add rope scaling support to MOSS-image by @kijai in #17172 [Partner Nodes] Renew partner-node sign-in tokens in API nodes by @dante01yoon in #16943 [Partner Nodes] feat(HeyGen): add HeyGen Voice

Article · llama.cpp. hexagon: add GGML_HEXAGON_VTCM_SIZE ( #31221 ) hexagon: add GGML_HEXAGON_VTCM_MB and GGML_HEXAGON_NTHREADS caps Change-Id: Idffe1613c8db6f96aa842d43d1b2efaed45462a8 hexagon: rename VTCM cap to GGML_HEXAGON_VTCM_SIZE, drop NTHREADS cap Change-Id: If4c5b9ce9fd6b725c3561421c2295d405

Article · Home Assistant. We’re detecting a new presence in the Works with Home Assistant program … and we’re pretty excited about it! A big welcome to our newest partner, Everything Smart Home 🎉 Founded by long-term community member Lewis Barclay and his dad, Dave Barclay, the family-run business brings

Article · Ollama. llama.cpp: version bump b11901 ( #19311 )

Article · llama.cpp. llama : enable moe amax rescaling for laguna ( #31147 ) llama : enable moe amax rescaling for laguna Assisted-by: Devin llama : declare laguna sm tensor support Website:

Article · Jellyfin. 🚀 Jellyfin Server 12.4 We are pleased to announce the latest stable release of Jellyfin, version 12.4! This minor release brings several bugfixes to improve your Jellyfin experience. As always, please ensure you take a full backup before upgrading! Discuss this release further on

Article · llama.cpp. opencl: write KV-cache views in set_tensor so a state restore is not dropped ( #30471 ) Website:

Article · llama.cpp. hexagon: MUL_MAT_ID row skipping optimizations and improvements ( #31169 ) hex-mmid: vectorize skipped row zeroing and added zeroing to fused MM_ID_NX hex-mmid: remove scalar div Website:

Article · llama.cpp. llama : move no_alloc out of llama_hparams ( #31141 ) llama : move no_alloc out of llama_hparams Assisted-by: dsh:llama.cpp/ggml-org/GLM-5.3-Flash-GGUF (2x DGX Spark (RDMA, TP=2)) cont : clean-up cont : clean-up Assisted-by: dsh:llama.cpp/ggml-org/GLM-5.3-Flash-GGUF (2x DGX Spark

Article · llama.cpp. models : replace ggml_tensor** out-params with output structs ( #31142 ) The hyper-connection and DSA graph builders wrote their extra outputs (post / comb gates, carried top-k selections, inject gates) through ggml_tensor** out-params. Return them in a small struct next to the p

Article · llama.cpp. ggml-cpu : rewrite concat as one row kernel for all types ( #31179 ) Every dst row comes from src0 or src1, or for dim 0 from the row of src0 followed by the row of src1, so a single kernel covers every type: a dense source row is one memcpy, quantized types included, and a strid

Article · llama.cpp. vulkan: rms norm optimization using subgroup reductions ( #29882 ) RMS Norm Subgroup reductions optimizaton Fix build error Improved optimization Make usage of BLOCK_SIZE consistent Add fast path to skip 2nd subgroupAdd Add AMD Vendor to subgroup reduction opt Add comment clarify

Article · llama.cpp. sycl: template the plain MMVQ launch triples ( #31013 ) Website:

Article · ServeTheHome. Today we are taking a look at ASRock's high-end Radeon AI inference card for workstations, the Creator Radeon AI Pro R9700. With 32GB of RAM and a $1700 price tag, the card is a highly affordable way to fill out a workstation for local AI inference The post ASRock Creator Radeon

Article · vLLM. [Bugfix][MRV2] Skip adaptive verification cost profiling in the CUDA …

Article · llama.cpp. sycl : bound reorder temp memory to 32 MiB ( #31062 ) Website:

Article · n8n. Merge branch 'master' into agent-1394-bug-builder-wastes-turns-on-nod…

Article · AnythingLLM. Built-in LLM for Docker AnythingLLM Docker now has a default local LLM experience right out of the box. The same kind of built-in engine that ships with AnythingLLM Desktop can now run inside the container : Lemonade , the fully open-source local inference server from AMD, runnin

Article · llama.cpp. sycl: template the per-quant concat impls ( #31008 )

Article · llama.cpp. sycl: add set_tensor_2d for faster tensor-split loading Pack host rows into chunks of up to 64 MiB when destination rows are contiguous, avoiding a separate upload call for each row. Keep the existing per-row copy path for non-contiguous destinations. On two Arc Pro B70s with gpt

Article · llama.cpp. tests: pin model resolution on real repo listings Extend test-model-resolution with 63 repo listings captured verbatim from the HF API, chosen to cover every file naming pattern found across 1237 GGUF repos of the major publishers and pruned to whole quant groups. Each repo pins

Article · llama.cpp. mtmd: add mtmd_helper_get_capabilities ( #31183 ) mtmd: add mtmd_helper_get_capabilities cap audio_stream add back old api for compat audio_gen_spk_ref, text_gen Website:

Article · noted.lol. Excerpt: Initiative gives communities a shared workspace for organizing projects and working together, all hosted on their own server.

Article · llama.cpp. ggml-cpu : fix unused variables in arm arch ( #31180 ) Website:

Article · llama.cpp. tests : refactor test-llama-archs to not link internal libllama symbols ( #30880 ) test-llama-archs included src/llama-arch.h and src/llama-model-saver.h, which prevented building on Windows with shared libraries since the used symbols are not exported from the DLL. declare llama

Article · llama.cpp. ci : fix race in webgpu memset ( #31162 ) Website:

Article · llama.cpp. metal : stop overflow from happening in large matrix output offsets ( #30072 ) Metal : stop overflow from happening in large matrix output offsets Assisted-by: Codex metal : widen batch offsets and MUL_MAT_ID token indices Assisted-by: Codex Co-authored-by: feni6 270037094+feni6@

Article · llama.cpp. deepseekv4.1: follow-up review fixes ( #31137 ) kv-cache : use ggml_backend_tensor_copy to refill the DSV4 K mirrors llama : move the Engram pad id to llama_hparams and the token map to llama_model The saver no longer needs the deepseek41 model class. deepseek41 : drop the F32 sr

Article · MarkTechPost. Tsinghua researchers released AKV, an SGLang-based serving system for cache engineering. LLM agents drop and reposition KV cache instead of summarizing history, cutting average cost 27.7% across 4 agent benchmarks. The post Tsinghua AI Open-Sources AKV, a Cache Engineering Server

Article · n8n. 1.123.88 (2026-10-26) Bug Fixes core: Release the SQLite write lock when the write connection fails to open ( #41918 ) ( 01b5e80 ) Update certifi to 2026.10.23 in task runner ( #41995 ) ( 3012503 )

Article · n8n. 2.44.7 (2026-10-26) Bug Fixes core: Start sub-workflow error workflow when a pre-run check fails ( #41978 ) ( 67d0e65 ) Email Trigger (IMAP) Node: Back off reconnects when the server drops each new connection ( #41927 ) ( 10a9518 ) Microsoft SharePoint Node: Fix the column mapper

Article · llama.cpp. cuda : use std::string instead of VLA for archName ( #30463 ) cuda : use std::string instead of VLA for archName Modernize the code (old C style) and fix the following warnings: ggml/src/ggml-cuda/ggml-cuda.cu:180:19: warning: variable length arrays in C++ are a Clang extension [

Article · llama.cpp. common, grammar: add recursive depth limit, make it a requirement ( #31153 ) common, grammar: add recursive depth limit, make it a requirement catch more recursive functions common : tidy up peg parser changes grammar : reject candidates in a loop, without a depth limit The depth

Article · llama.cpp. tests : adjust default threads for test-llama-context ( #31157 ) Website:

Article · llama.cpp. qwen: merge tensors on convert ( #31028 ) qwen: merge tensors on convert convert: use generator functions qwen4exp: remove conts and pre-scale Website:

Article · Ollama. Speech Ollama can now generate speech on macOS. ollama run plays text or a chat model's reply aloud as the audio is generated, and apps can stream audio through the OpenAI-compatible /v1/audio/speech endpoint. ollama run phonon " Hello from Ollama. " d3 decision models d3 brings

Article · llama.cpp. llama: fix exhausting CPU and MEM caused by dry sequence breaker ( #31035 ) Website:

Article · llama.cpp. llama-quant : fix 4D tensor quantization using ggml_nrows() and name-based expert detection ( #30820 ) llama-quant : fix quantizing 4D tensors nrows_total was ne[1]*ne[2], so for a 4D tensor only the first ne[2] slices were quantized and the rest of the rows were dropped. Use ggm

Article · noted.lol. VoidAccess is an MIT-licensed threat research platform with Docker Compose, Tor integration, and locally hosted AI options.

Article · llama.cpp. grammar: limit recursion depth to prevent stack overflow ( #29495 ) ( #29755 ) Website:

Article · llama.cpp. llama : respect no_alloc in all memory caches ( #31087 ) llama : respect no_alloc in memory caches Assisted-by: dsh:llama.cpp/ggml-org/DeepSeek-V4-Flash-Vision-Exp-GGUF (Mac Studio, M2 Ultra, 192GB) llama : TODO: move no_alloc out of llama_hparams Assisted-by: dsh:llama.cpp/ggml-

Article · llama.cpp. tools : reject truncated GGUF tensor data in merge ( #31044 ) Co-authored-by: Shreya Agrawal shreyaagrawal@Shreyas-MacBook-Air.local Website:

Article · llama.cpp. CUDA: fuse FFN gate/up matmuls and GLU in MMQ with interleaved gate/up rows ( #29948 ) cuda: fuse FFN gate/up and GLU in MMQ with interleaved gate/up rows For prompt processing, run the gate and up matmuls of a dense FFN as one mul_mat_q launch and apply the GLU in the write-back

Article · llama.cpp. ggml-cuda: fuse qwen MRoPE + set-rows ( #31025 ) Website:

Article · llama.cpp. vulkan: Add a larger tile size for conv3d (coopmat2 only) ( #29920 ) vulkan: Add a larger tile size for conv3d (coopmat2 only) This adds a larger tile size, in practice it is only for coopmat2 where we don't need shared memory to store the output (would use more than 64KB). Add s

Article · llama.cpp. vulkan : one warp tile per subgroup in the small matmul tiles ( #30106 ) WM of the small tiles was 32, or 8 for 8-wide subgroups, whatever the tile's BLOCK_SIZE and WARP. With 16-wide subgroups (Mali) the second subgroup of a workgroup wrote into the next tile's columns. With 8-w

Article · llama.cpp. hexagon: generic dims collapse/reshape for unary, binary and matmul ops ( #31071 ) hex-tools: support for --compare option in the profile.py hex-unary: improve collapse logic to also do splits if nrows is small hex-colapse: handle all colapsing at the host and make it generic hex

Article · llama.cpp. opencl: add topk and lightning indexer ( #31026 ) opencl: add topk opencl: optimize topk opencl: add fp lightning indexer opencl: optimize lightning indexer opencl: refactor opencl: refactor tiled topk kernel opencl: adjust format Website:

Article · noted.lol. Miss flipping through TV channels? Tunarr lets you build your own live television network using the media already on your server.

Article · llama.cpp. opencl: add initial q2_k support ( #30999 ) opencl: add mul_mv_q2_k_f32 opencl: add q2_k SOA mv and tiled mm opencl: add q2_k noshuffle gemm/gemv for adreno opencl: add q2_k dp4a gemm for adreno Co-authored-by: Hongqiang Wang wangh@qti.qualcomm.com opencl: tune q2_k kernels for a

Article · LiteLLM. Verify this release cosign verify \ --key

Article · LiteLLM. Verify this release cosign verify \ --key

Article · LiteLLM. Verify this release cosign verify \ --key

Article · LiteLLM. Verify this release cosign verify \ --key

Article · Open WebUI. Added 🎯 Load selected skills on demand. Administrators can turn on "Load selected skills on demand" under Admin Settings > General or set "ENABLE_SKILL_LAZY_LOADING", so skills selected in the chat or attached to a model are only listed for the model to load when it needs them in

Article · llama.cpp. chat : reject JSON nested more than 128 levels before walking it ( #30242 ) chat : reject JSON nested more than 128 levels before walking it The schema builder, the jinja engine and the decision code walk request JSON with recursion. A request nested deep enough overflows the sta

Article · MarkTechPost. Microsoft Research open-sourced Agensh, a multi-agent harness with no central orchestrator. Self-organizing workers share a Git repo, chat and a findings board, scaling from 1 to 1,024 agents on ProgramBench under a 6-hour budget. The post Microsoft Open-Sources Agensh: A Multi-A

Article · MarkTechPost. Cognition released the Devin Vulnerability Fixes plugin, 3 MIT-licensed agent skills that inventory, patch and test PHP, Node.js and self-managed GitLab deployments hit by this week’s critical RCE flaws, then open 1 PR per repo. The post Cognition Open-Sources an Agent Plugin tha

Article · MarkTechPost. Explore a complete coding guide comparing Underdog Saluki 27B with ISTA's GSQ-RCO quantization, examining GGUF headers, memory usage, and BFCL tool-calling performance in stock llama.cpp The post How to Run Underdog Saluki 27B: 2-Bit Qwen3.8-27B Tool-Calling Guide appeared first

Article · llama.cpp. model: DSV4.1 flash ( #31086 ) model : add DeepSeek-V4.1-Flash (deepseek41) deepseek41 : remove MXFP8, convert the Engram table to BF16 deepseek41 : compute the compressed-KV read layers once in the cache deepseek41 : use get_cells() for the shared-history check, drop llama_kv_ca

Article · llama.cpp. jinja : fix select and selectattr argument handling ( #31059 ) jinja : fix select and selectattr argument handling jinja : address review feedback jinja : Simplified test attr Website:

Article · llama.cpp. hexagon: KV-head team split for multicore HMX flash attention ( #31055 ) hexagon: KV-head team split for HMX flash attention in multicore hexagon: merge FA_HEAD_SPLIT and FA_SPLIT into single GGML_HEXAGON_FA_SPLIT flag 0 = off (token-split only); 1 = head-split, plus team-split w

Article · llama.cpp. server: refactor server entrypoint ( #31113 ) server_role cont Website:

Article · llama.cpp. metal : remove mul_mm_id amax rescaling ( #30924 ) The src1 power-of-two rescale in kernel_mul_mm_id is superseded by the per expert-token amax scaling done in the compute graph (ref: #30861 ), which keeps the MoE down-projection activations within f16 range before the op runs. D

Article · llama.cpp. rpc : add bounds check in rdma_recv ( #30399 ) Website:

Article · llama.cpp. minimax-m3 : forward quant scales to the graph (fixes #31063 ) ( #31097 ) The generic scale pass in llama-model.cpp loads the per-expert NVFP4 scale2 tensors (ffn_{up,gate,down}_exps_s) and the per-tensor MXFP8 scales for the dense FFN and the shared expert, but minimax-m3.cpp ne

Article · llama.cpp. tests : fix test-llama-context for large n_batch ( #31083 ) Website:

Article · llama.cpp. llama : deny -sm tensor for arches with moe amax rescaling ( #31081 ) The amax POOL_1D reduction spans the whole matmul row, so a device cannot compute it from an axis-0 split. The Meta split-state case reported MIRRORED, which gives each device a partial max of its own slice. Dr

Article · llama.cpp. ggml : allow MUL_MAT_ID -1 index to skip the computation ( #26631 ) ggml : allow MUL_MAT_ID -1 index to skip the computation nits update vulkan code ggml-webgpu: ensure all threads reach workgroup barriers in mul_mat_id_vec (cherry picked from commit d8a4b9e ) fix cuda openvino e

Article · Ollama. Flint on Linux and Windows Flint now runs on Linux and Windows as well as macOS. Other improvements Streaming OpenAI Responses and Anthropic Messages requests now report generation failures instead of ending silently after partial output. ollama launch now passes cloud models' co

Article · noted.lol. A self-hosted dead man’s switch that helps solo admins leave a plan for the people who may need to take over.

Article · llama.cpp. tests : add speculative rejection draft sampling test ( #30213 ) tests : add speculative rejection draft sampling test This commit adds a test for speculative rejection sampling test that excercises the speculative rejection sampling mechanism in common/speculative.cpp. The motiv

Article · llama.cpp. vulkan: ggml src1 precision support ( #30656 ) vulkan: update accumulator precision selection handle fp32 overflows through scaling handle overflows in copy-to-fp16 activations add missing cases simplify code add tests add ggml prec get helpers, use in Vulkan backend clean up com

Article · noted.lol. Follow blogs and YouTube channels together with WyrmRSS, a self-hosted RSS reader with flexible feeds and a mobile-friendly layout.

Article · llama.cpp. HIP: add byte permute fallback when __builtin_amdgcn_perm is unavailable ( #30915 ) HIP: add byte permute fallback when __builtin_amdgcn_perm is unavailable amdgcnspirv rejects target-specific builtins unless they are guarded by __builtin_amdgcn_is_invocable , so 66 of 145 HIP fi

Article · LiteLLM. Important The image-signing key has changed. Starting with this release, every Docker image and Helm chart is signed with the new cosign key from commit e82a3c6 . Verifying with the previous key from commit 0112e53 will fail. Update anything that pins the old cosign.pub : cosign

Article · LiteLLM. Important The image-signing key has changed. Starting with this release, every Docker image and Helm chart is signed with the new cosign key from commit e82a3c6 . Verifying with the previous key from commit 0112e53 will fail. Update anything that pins the old cosign.pub : cosign

Article · LiteLLM. Important The image-signing key has changed. Starting with this release, every Docker image and Helm chart is signed with the new cosign key from commit e82a3c6 . Verifying with the previous key from commit 0112e53 will fail. Update anything that pins the old cosign.pub : cosign

Article · LiteLLM. Important The image-signing key has changed. Starting with this release, every Docker image and Helm chart is signed with the new cosign key from commit e82a3c6 . Verifying with the previous key from commit 0112e53 will fail. Update anything that pins the old cosign.pub : cosign

Article · llama.cpp. mtmd: move build_inp_temporal_merge to base class ( #31065 ) Website:

Article · llama.cpp. mtmd : lfm2 resize matches torchvision ( #30943 ) mtmd : lfm2 bilinear resize matches torchvision Lfm2VlImageProcessor and d1-omni vision.py resize uint8 images with torchvision (antialias=True). That is Pillow's algorithm, except that the weights are int16 with as many fractiona

Article · llama.cpp. mtmd : add Soprano TTS support ( #28607 ) Add Soprano-80M and Soprano-1.1-80M (

Article · MarkTechPost. NVIDIA released Nemotron 3.5 Bolt E4B 4GB, an open small language model post-trained from Gemma 4 E4B for local voice assistants, game characters and robots, targeting about 4GB of VRAM on RTX and Jetson. The post NVIDIA Releases Nemotron 3.5 Bolt E4B 4GB: An Open Small Language

Article · llama.cpp. hexagon: add SGN op ( #30346 ) hexagon : add SGN support Assisted-by: OpenCode hexagon : optimize SGN with sign bit Assisted-by: OpenCode hex-sgn: add f16 hex-unary: update SGN to use generic unary_colapse and relax colapse checks to apply to more conditions hex-unary: improve di

Article · llama.cpp. common : allow concurrent downloads across processes ( #29312 ) Signed-off-by: Adrien Gallouët angt@huggingface.co Website:

Article · llama.cpp. hip: fix bf16 WMMA operand vector type for gfx11/gfx12 ( #30885 ) The __builtin_amdgcn_wmma_f32_16x16x16_bf16_w32 (and _gfx12) builtins take their A/B operands as vectors of short (_ExtVector / _ExtVector ), not __bf16. Declaring the fragments as ext_vector_type(N) __bf16 fails t

Article · llama.cpp. ggm-cuda: Enforce DP4A MMQ tile height multiple of warp size ( #30837 ) Added static assertion to enforce DP4A MMQ tile height requirement. Website:

Article · Ollama. mlxrunner: import and list Phonon catalog voices docs: document Phonon voices

Article · Simon Willison. Kimi K3.1 Pro landed on the Kimi API today - it's not yet open weights, but Moonshot promise "Full model weights will be released by Nov 4, 2026". They're also offering a "Highspeed" version that's 2-3x faster, at twice the price (except for input token cache hits, which are pric

Article · llama.cpp. CUDA: fix build after chunked CUDA graphs ( #30804 ) Apply suggestion from @JohannesGaessler Co-authored-by: Johannes Gäßler johannesg@5d6.de Fix newlines in common.cuh Co-authored-by: Johannes Gäßler johannesg@5d6.de

Article · llama.cpp. allows the usage of 4-10 gpus for stepfun and laguna in tensor paral…

Article · llama.cpp. qwen4exp : build the QSA selection without the kq_mask clean lang qwen4exp : reshape the fused QSA selection to the kq_mask streams ggml_top_k_mask with a single keep row returns [n_kv, n_tokens], while build_attn_qsa expects the kq_mask layout [n_kv, n_tokens/n_stream, 1, n_stre

Article · Ollama. NVIDIA Nemotron 3.5 Bolt E4B Nemotron 3.5 Bolt E4B is NVIDIA’s smaller, Gemma 4-based vision-language model for fast local conversations, image understanding, and tool use. It supports a 128K-token context window. The model includes a DFlash draft for speculative decoding. ollama

Article · ServeTheHome. We take a look at the MSI XpertStation WS300, the company's top local AI solution powered by the NVIDIA DGX Station platform The post MSI XpertStation WS300 Review NVIDIA DGX Station GB300 appeared first on ServeTheHome .

Article · llama.cpp. hexagon: add CEIL op ( #31006 ) hex-ceil: add CEIL(f32) op hex-ceil: add f16 version hex-ops: update supported ops hex-ops: update docs with supported ops Website:

Article · llama.cpp. CUDA: chunk graphs to amortize graph launch ( #30804 )

Article · LocalAI. Patch release v4.12.1. Security fix only. Security Hardened the external URL policy for user-supplied URLs (chat vision image/audio/video URLs, image and audio endpoints). Redirect targets are now validated on every hop, 100.64.0.0/10 and 0.0.0.0/8 are treated as non-public, and

Article · Ollama. renderers: add gemma4-nemotron renderer Add a Gemma 4 renderer variant that renders assistant content before its tool calls rather than after the tool responses. Models opt in with RENDERER gemma4-nemotron; gemma4 auto-detection is unchanged. Adds the corresponding chat template

Article · llama.cpp. ggml : fix time overflow in ggml_time_us ( #30969 ) fix time overflow use double Website:

Article · noted.lol. Noted Giveaways is here! Discover handpicked homelab and PC hardware giveaways, all manually reviewed and approved. No endless lists of junk promotions designed to harvest your email. Just worthwhile opportunities to win gear for your next build.

Article · selfh.st. Shell shocked

Article · noted.lol. Edi Life OS unites habit tracking, goals, Kanban, finance, and an optional Model Context Protocol AI server into a self-hosted dashboard.

Article · llama.cpp. CANN: missing const qualifiers added for consistency ( #30933 ) Website:

Article · n8n. 1.123.87 (2026-10-23)

Article · llama.cpp. sycl : pass device into mul_mat_uses_onednn ( #31014 ) Website:

Article · n8n. 2.45.3 (2026-10-23) Bug Fixes editor: Increase light-mode text contrast ( #41807 ) ( 9f41c0f )

Article · Dify. Dify 1.18.0 brings Agents that route workflows , portable app packages , deeper visibility into nested workflow runs , and the opt-in Agent Knowledge Base experience. OpenAPI and the new difyctl expand how you build and manage apps from the command line. Important Read the upgrad

Article · llama.cpp. spec : add probabilistic sampling for DFlash2 drafts spec : apply review suggestions Co-authored-by: Pranesh Gonegandla pgonegandla@nvidia.com

Article · llama.cpp. vocab : fix Nemotron-3.5-Embed tokenization Update src/llama-vocab.h Co-authored-by: Sigbjørn Skjæret sigbjorn.skjaeret@huggingface.co rename Update src/llama-vocab.cpp Co-authored-by: Daniel Bevenius daniel.bevenius@gmail.com Co-authored-by: Sigbjørn Skjæret sigbjorn.skjaeret@hu

Article · llama.cpp. hexagon: Op Fusion refactoring, optimizations and new fusions like `B…

Article · llama.cpp. sycl: Replace host waits in fattn-mkl with explicit dependencies ( #30 …

Article · llama.cpp. Signed-off-by: Todd Malsbary todd.malsbary@intel.com

Article · Jellyfin. 🚀 Jellyfin Server 12.3 We are pleased to announce the latest stable release of Jellyfin, version 12.3! This minor release brings several bugfixes to improve your Jellyfin experience. As always, please ensure you take a full backup before upgrading! Discuss this release further on

Article · ComfyUI. Article from ComfyUI.

Article · llama.cpp. server: Substitute rerank placeholder from the original template only. ( #30864 ) Website:

Article · llama.cpp. server: realtime API: send complete after committed audio drains ( #30994 ) server: realtime API: send complete after committed audio drains server: realtime API: complete a commit at its end row Completion is triggered when the stream runs dry, which is a timing proxy: audio app

Article · Ollama. Replace the prerelease checkpoint path with Gradium's GA format, including default-voice import and speech conditioning. Set the 0.40.9 minimum, update the speech docs, and test import, inference setup, and rejection of the old checkpoint.

Article · Paperless-ngx. paperless-ngx 3.3.4 Note This security maintenance release updates dependencies to address the eMudhra / emSign CA certificate security incident and security vulnerabilities in Pillow and aiohttp. These issues affect third-party dependencies rather than Paperless-ngx’s own code.

Article · Hugging Face. Article from Hugging Face.

Article · llama.cpp. opencl: add depthwise conv2d ( #30900 ) opencl: add Adreno depthwise conv2d Assisted-by: OpenCode opencl: use local program for depthwise conv2d Assisted-by: OpenCode opencl: use singed type for args Co-authored-by: Li He lih@qti.qualcomm.com Website:

Article · llama.cpp. mtmd : add MiniCPM-OCR 1.0 support ( #30979 ) Signed-off-by: tc-mb tianchi_cai@icloud.com Website:

Article · Ollama. flint: fix replies with thinking disabled Close the thinking section when think:false so Flint answers directly. Default and think:true prompts are unchanged. update template

Article · Simon Willison. News today from Anthropic: We’re committed to keeping Agent Skills open, community-driven, and vendor-neutral. Today, we are donating the standard to the Agentic AI Foundation (AAIF) under the Linux Foundation,

Article · llama.cpp. glm5 : optimize inputs ( #30932 ) glm5 : avoid sel_mask_2d input cont : avoid new_pool_flat input Website:

Article · llama.cpp. model : fix qwen4exp layer inputs for DFlash drafters ( #30927 ) qwen4exp published the hc-wide residual [n_embd, hc, n_tokens] as t_layer_inp, but the drafter reads layer inputs as [n_embd, n_tokens] rows: llama_context sizes each embd_layer_inp buffer as n_embd * n_batch and th

Article · llama.cpp. Multi-column on Intel XMX matrix engines refactored ( #29864 ) Multi-column on Intel XMX matrix engines refactored Multi-column (2 to 80 activation columns) mul_mat_vec_q for reordered K-quant and Q8_0 weights on the Intel XMX matrix engines (int8 DPAS, ESIMD), for the SYCL backe

Article · noted.lol. You've probably got years of photos, videos, and music sitting on your NAS that you rarely revisit. Spooly gives that forgotten media a second life with swipeable discovery, personal radio stations, and a fresh way to explore your collection.

Article · LocalAI. 🎉 LocalAI 4.12.0 Release! 🚀 LocalAI 4.12.0 is out! This release redesigns the web UI and adds streaming voice responses, native decision routing, and more control over model memory and prompt state. It also adds image upscaling, live motion capture, Qdrant stores, and authenticat

Article · llama.cpp. Article from llama.cpp.

Article · llama.cpp. nemotron : support non-standard GQA, BPE tokenizer and domyn-small pre-type ( #23343 ) nemotron : support non-standard GQA, BPE tokenizer and domyn-small pre-type gguf-py : write Metaspace prepend_scheme as GGUF metadata vocab : parse metaspace prepend_scheme into enum, rename ke

Article · llama.cpp. vulkan: replace std::aligned_alloc with posix_memalign for termux com…

Article · llama.cpp. server : support typed content input for RERANK endpoint (

Article · n8n. 1.123.86 (2026-10-22) Bug Fixes deps: bump sharp to 0.35.5 on 1.x (DEVP-1241) ( 9fa95f0 )

Article · llama.cpp. sycl: fix CONV_2D wrong results for non-contiguous input (

Article · n8n. 2.45.2 (2026-10-22) Bug Fixes core: Read type availability policies from memory at enforcement points ( #41621 ) ( 6433596 ) Postgres Node: Prevent global pgtypes overwrite ( #41666 ) ( eff4a45 ) Update logos in credential oauth confirmation screen and instance emails ( #41637 )

Article · n8n. 2.44.6 (2026-10-22) Bug Fixes Postgres Node: Prevent global pgtypes overwrite ( #41667 ) ( 654315e )

Article · llama.cpp. vocab : add clean_spaces GGUF key ( #30940 ) Website:

Article · llama.cpp. vocab : use Tekken pre-tokenizer for pixtral rather than the Llama 3 one ( #30960 ) Website:

Article · llama.cpp. models : Record layer inputs for all granite models ( #30957 ) Branch: GraniteSpeculatorSupport AI-usage: full (Claude + Sonnet 5.5) Signed-off-by: Gabe Goodhart ghart@us.ibm.com Website:

Article · llama.cpp. kleidiai : add Q8_0 SVE dotprod GEMV and SVE i8mm GEMM kernels ( #30808 ) Website:

Article · Immich. Caution This version contains a fix for today's CA compromise (

Article · llama.cpp. sycl: merge the fp16 and fp32 dequant switches ( #30740 ) Website:

Article · llama.cpp. webgpu : add Q2_0 support ( #30914 ) webgpu : add Q2_0 row dequantization webgpu : add Q2_0 matrix multiplication Website:

Article · llama.cpp. dflash : load optional FC input scale ( #30945 ) Assisted-by: Codex Website:

Article · llama.cpp. lora : support prism.hadamard folded weights ( #30910 ) lora : support prism.hadamard folded weights tests : remove LoRA test from test-llama-archs Website:

Article · llama.cpp. opencl: add F32-to-I32 ARGMAX ( #30897 ) opencl: add F32-to-I32 ARGMAX Assisted-by: OpenCode opencl: init to 0 to be consistent with cpu Co-authored-by: Li He lih@qti.qualcomm.com Website:

Article · Ollama. llama.cpp update ( #19194 )

Article · TLDR AI. Article from TLDR AI.

Article · ComfyUI. Article from ComfyUI.

Article · llama.cpp. models : add Flint model (FlintForCausalLM) ( #30928 ) Add Flint model (FlintForCausalLM). Assisted by Codex cleanup one more Co-authored-by: Sigbjørn Skjæret sigbjorn.skjaeret@huggingface.co Website:

Article · llama.cpp. jinja : fix namespace attributes shadowed by dict methods ( #30941 ) Website:

Article · MarkTechPost. Vercel released durable-harness, an open-source package that logs every model and tool call so Pi, OpenCode, Claude Code and Codex sessions resume after crashes, timeouts and redeploys on Vercel, Postgres or a laptop. The post Vercel Open-Sources durable-harness for Crash-Proof C

Article · llama.cpp. model: Add support for SenseNova U1 model ( #28919 ) model : add SenseNova U1 understanding support Add support for the dense SenseNova U1 understanding branch, including the U1.5-8B-MoT model. Add HF-to-GGUF conversion for NEOChatModel, excluding MoT generation tensors and conca

Article · Ollama. Recursive Flint Flint brings Recursive's coding model to Ollama. It is designed for the work coding agents do across a repository: understanding unfamiliar code, tracing bugs, implementing features, and adding tests. Reasoning, multi-turn tool use, and a 262K-token context window

Article · llama.cpp. llama: create new fused op for QSA mask ( #30640 ) llama: create new fused op for QSA scoring enroll glm5.3-next too tests : initialize the remaining tensors in test_kpool_mask The harness sentinels were left uninitialized, so leftover memory could read as NaN and fail the NaN ch

Article · Ollama. We were still using a pre-release, and the download site is flaky, so lets get current on the latest CUDA Toolkit for WoA support.

Article · llama.cpp. sycl: q8 fast dequant ( #30775 ) Website:

Article · noted.lol. Built for teams that need to keep feedback in-house, RetroGemini brings guided retrospectives, anonymous voting, and health checks to your server.

Article · llama.cpp. jinja: support hexadecimal string escapes ( #30929 ) Website:

Article · llama.cpp. ggml-opencl : add IM2COL_3D support ( #30895 ) ggml-opencl : add IM2COL_3D support Assisted-by: OpenCode opencl: use same type for kernel and host opencl: fix for older adreno Co-authored-by: Li He lih@qti.qualcomm.com Website:

Article · Ollama. Preserve source tool JSON alongside decoded API values so Flint can retain numeric types and schema order in prompts. Existing API marshaling is unchanged. Assemble ModelOpt expert banks during safetensors import, and reject direct Flint NVFP4 creation before conversion. Other mo

Article · n8n. 2.45.1 (2026-10-21) Bug Fixes core: Claim an idle worker's next queue job and its lock in one step ( #41558 ) ( 5f0b562 ) editor: Center badge text when icons are absent ( #41581 ) ( 648fa2a ) Microsoft Teams Trigger Node: Renew the Graph subscription via lifecycle notifications

Article · llama.cpp. mtmd: add state object for mtmd_gen ( #30902 ) Website:

Article · llama.cpp. server : forward llama.cpp sampling params on /v1/messages ( #30856 ) server : forward llama.cpp sampling params on /v1/messages server : pass the request through on /v1/messages instead of copying listed keys NIT: Change comment from multiline to single line server : accept llam

Article · llama.cpp. llama : optimize vocabulary loading ( #30679 ) llama : optimize vocabulary loading Update src/unicode.h Co-authored-by: Georgi Gerganov ggerganov@gmail.com CR : added reserve in gguf_kv_to_str CR : trailing space fix Co-authored-by: Georgi Gerganov ggerganov@gmail.com Website:

Article · llama.cpp. llama : place d2t by its set_rows use so it can stay on the GPU ( #30912 ) Website:

Article · llama.cpp. metal : fix memory leaks and cleanup logic in GPU capture path ( #30890 ) Website:

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . Every release is signed with the same key introduced in commit 0112e53 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way

Article · llama.cpp. hexagon: neon-vectorized and threaded repack (tiling) ( #30901 ) add vectorized q4_0 repack apply to more types hex-repack: neon-vectorize the rest of the types q1/q2/f16/... hex-repack: add simple chunking / threading Co-authored-by: hubezhou hubezhou@qti.qualcomm.com Website:

Article · noted.lol. Quomodo brings a central web interface to Podman stacks, with Git-based deployments and rootless container support.

Article · llama.cpp. opencl: add dp4a GEMM variants to vector load activation ( #30753 ) opencl: add dp4a GEMM variants with vector activation loads opencl: rename kernels and env vars Website:

Article · llama.cpp. sycl : fix fa crash for Query head dimension mismatch (neq0 != HSK / D=2048/4096) ( #30286 ) Website:

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . Every release is signed with the same key introduced in commit 0112e53 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way

Article · llama.cpp. sycl: collapse duplicate need_check launch bodies in mmq ( #30739 ) Website:

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . Every release is signed with the same key introduced in commit 0112e53 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way

Article · llama.cpp. ggml-cpu : compute f32 gelu_quick and geglu_quick without the fp16 table ( #30737 ) ggml-cpu : compute f32 gelu_quick and geglu_quick without the fp16 table ggml-cpu : keep the GGML_GELU_QUICK_FP16 table path ggml-cpu : add the VXE gelu_quick path tests : add test-gelu-quick for

Article · n8n. 2.44.5 (2026-10-21) Bug Fixes core: Key queue job outcomes by job on shutdown resume ( #41404 ) ( 1f074d4 ) editor: Load credential types on startup ( #41405 ) ( 6ff3198 )

Article · LiteLLM. fix(llms): sanitize trailing slashes in provider get_complete_url test(llms): move trailing-slash url tests into each provider's mapped unit file fix(llms): strip trailing slashes in Perplexity search and Ollama embedding URLs and run the Topaz tests in CI fix(llms): strip traili

Article · llama.cpp. llama : scale moe down projection per expert token ( #30861 ) llama : scale moe down projection per expert token Replace the global amax rescaling for the MoE down projection with a per-expert-token factor computed by a POOL_1D max reduction along the intermediate dimension. Undo

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . Every release is signed with the same key introduced in commit 0112e53 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way

Article · llama.cpp. Sycl iq3 crash fix ( #30604 ) Website:

Article · llama.cpp. server : fix usage.completion_tokens undercount with n > 1 ( #29486 ) Assisted-by: Copilot Website:

Article · llama.cpp. server : add POST /tts endpoint for TTS models ( #26603 ) mtmd: add audio out stream api add /tts endpoint add docs wire up mtmd_helper_model_can_chat upload speaker_ref via form-data expose seed to api fix header for pcm wip fix penalty_last_n fix server_output_limits fix pocket

Article · llama.cpp. server: add realtime transcription API ( #30636 ) server: add realtime transcription API drop transcription_sessions proxy ws add docs server: let the API key middleware guard the websocket httplib runs the pre-routing handler before the websocket upgrade, so the API key middlewa

Article · Ollama. What's Changed pplx-decider is a 27B decision model from Perplexity, fine-tuned from Qwen3.8-27B Continuing an OpenAI-compatible Responses conversation no longer fails when switching between models with and without thinking, or when the history contains another provider's compact

Article · llama.cpp. model : support Nemotron-3(.5)-Embed ( #30877 ) Website:

Article · LiteLLM. Verify this release cosign verify \ --key

Article · llama.cpp. qwen: reshape output to 2d instead of 3d ( #30823 ) qwen: reshape output to 2d instead of 3d remove final reshape, fix comments Website:

Article · llama.cpp. server: Support logprobs with speculative decoding ( #27196 ) return logprobs in spec path server: add regression test for logprobs with speculative decoding Asserts accepted draft tokens report the target model's logprobs, comparing a draft-enabled run against a no-draft baselin

Article · llama.cpp. mtmd : lfm2 tiles and layout match transformers ( #30261 ) mtmd : lfm2 tiles stretch the image to the grid Lfm2VlImageProcessor.crop_image_to_patches resizes the image straight to (512 * rows, 512 * cols) before it cuts the tiles. llama.cpp kept the aspect ratio and added black b

Article · llama.cpp. model : add H2O-Lightning-4B decision model ( #30361 ) model : add H2O-Lightning-4B decision model Adds the h2o-lightning decision type for h2oai/h2o-lightning-4b (Qwen3.5-4B with a merged decision LoRA). None of the existing types match this model. It answers with " A" rather th

Article · llama.cpp. mtmd: fix gemma 4 ffn act ( #30882 ) Website:

Article · llama.cpp. ggml-cuda : support ssm_conv kernel size 7 ( #30873 ) Website:

Article · llama.cpp. server: Router wasn't propagating the client disconnects fully to the child, the response wait could cause a delayed cancellation ( #29707 ) Router wasn't propagating the client disconnects fully to the child, the response wait could cause a delayed cancellation tests: satisfy ro

Article · llama.cpp. server: fix missing output_index in Responses API SSE events for compatibility ( #30863 ) server: include output_index in Responses API SSE events server: reuse Anthropic streaming code for fix output_index server: remove unnecessary generated_tool_call_ids and oai_resp_tool_call

Article · llama.cpp. mtmd : read LFM2 tiling params from GGUF metadata ( #30840 ) mtmd : read LFM2 tiling params from GGUF metadata Write min_tiles, max_tiles, tile_size and max_pixels_tolerance from processor_config.json when converting LFM2-VL mmproj, read them in clip.cpp with the previous constan

Article · llama.cpp. Newer versions of Clang diagnose an incompatible pointer type error when assigning a pointer to float to a pointer to half.

Article · llama.cpp. model : drop the redundant precision overrides in the Kolibri-1 router ( #30865 ) model : drop the redundant precision overrides in the Kolibri-1 router Follow-up to the review in #30673 : the router matmul works with the default precision, so both ggml_prec_set_* calls can go aw

Article · llama.cpp. opencl: add several gemm variants for small batches and some fixes ( #30850 ) opencl: fix OOB read in mul_mv_mxfp4_f32_flat opencl: remove large-N guard that forces CPU weight placement this condition claims that large batch produces incorrect results on adreno and cannot be repr

Article · ComfyUI. Article from ComfyUI.

Article · llama.cpp. ggml-cpu : select the NEON fp32->fp16 conversion ( #30505 ) The _ ARM_NEON branch defines both GGML_CPU_COMPUTE * macros but selects only FP16_TO_FP32, so FP32_TO_FP16 falls through to the portable software conversion while neon_compute_fp32_to_fp16, a single fcvt, stays defined

Article · llama.cpp. HIP/CUDA: Fix reduction in mmf getting tile widths not devisiable by 64 ( #30836 ) Website:

Article · llama.cpp. ggml-zdnn: introduce flash attention op ( #30512 ) ggml-zdnn: add scratch buffer pool Signed-off-by: Aaron Teo aaron.teo1@ibm.com ggml-zdnn: implement fattn, ditch scratch slots in favour of scratch buffer pool Signed-off-by: Aaron Teo aaron.teo1@ibm.com docs: add note about fatt

Article · llama.cpp. refactor : move LFM2 shortconv and FFN blocks into a shared LFM base …

Article · Ollama. Deduplicate MLX CUDA runtime payloads Move CUDA libraries shared by the GGML and MLX backends into the sibling CUDA payload, and teach in-process and subprocess MLX loading to resolve that packaged dependency directory. Keep MLX-private libraries and JIT headers in the MLX overla

Article · ComfyUI. Article from ComfyUI.

Article · llama.cpp. CUDA: Clear CUDA graphs after legacy memory pool flush ( #26771 ) Assisted-by: Claude Opus 5.5 Website:

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . Every release is signed with the same key introduced in commit 0112e53 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . Every release is signed with the same key introduced in commit 0112e53 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way

Article · llama.cpp. hexagon: fix VTCM size query on images that need layout out-params (#…

Article · llama.cpp. server : avoid repeated mmproj memory margin addition on sleep/resume…

Article · noted.lol. Petal lets remote collaborators share movable desktop windows and run their own media server, although setup is still hands-on.

Article · MarkTechPost. DeepSeek has open-sourced DeepInfer, the Apache 2.0 inference framework that serves DeepSeek-V4.1-Flash in production. It targets NVIDIA SM100 GPUs and Huawei Ascend 950 NPUs, and ships offline inference only. The post DeepSeek Open-Sources DeepInfer, the Inference Framework Behi

Article · llama.cpp. server : pass id_slot through the Anthropic /v1/messages conversion ( #28554 ) server : pass id_slot through the Anthropic /v1/messages conversion server_chat_convert_anthropic_to_oai copies a small whitelist of common params from the Anthropic request into the internal OAI-style

Article · llama.cpp. metal : optimize argmax and pool_1d Add a two-pass partial reduction for large-row argmax and use cooperative threadgroup reductions for pool_1d. Allocate scratch space for the argmax partials through the Metal buffer allocator. Assisted-by: pi:llama.cpp/ggml-org/DeepSeek-V4-Flas

Article · llama.cpp. sycl : replace memset due to not always ordered ( #30812 )

Article · llama.cpp. spec : save the dynamic ngram cache file save the cache at the common_speculative_impl_ngram_cache destruction remove save_static/save_dynamic flags as well as the relevant TODO if the dynamic cache file does not exist, start with an empty cache that is written on exit. if it exi

Article · llama.cpp. metal : fix missing MoE kernel variant for 9 experts per token ( #30610 ) ( #30699 ) Website:

Article · Immich. Turns out the last people performance fix wasn't quite there for some people. We have now finally resolved that issue. What's Changed 🐛 Bug fixes fix: missing translation by @waclaw66 in #32485 perf(server): more efficient person queries by @mertalev in #32523 Full Changelog : v3

Article · n8n. 2.44.4 (2026-10-19) Bug Fixes core: Reuse connections in the log streaming webhook destination ( #41281 ) ( b79afe2 ) core: Turn off expression spans by default ( #41309 ) ( 6965193 )

Article · n8n. 2.45.0 (2026-10-20) Bug Fixes AI Agent Node: Replay Anthropic thinking turns with their tool arguments ( #41241 ) ( ea98bd0 ) Align data table and variable editing on protected instances ( #41258 ) ( d4bde14 ) Allow skipping table initialization ( #40437 ) ( d245b17 ) API: Apply

Article · llama.cpp. vocab : add \p{M} to lookaheads in Tekken and GPT-4o pre-tokenizer regexes ( #30845 ) Website:

Article · llama.cpp. tests : make MoE router selection independent of the backend noise ( #30818 ) tests : make MoE router selection independent of the backend noise The router scores are in the range [0, 1), so a random draw can leave the last expert (or expert group) still selected and the first on

Article · MarkTechPost. Cloudflare has previewed Clairic, a System One extraction model on Workers AI that returns typed values with confidence scores in one forward pass. It costs $0.10 per million input tokens and charges nothing for output. The post Cloudflare Releases Clairic, a System One Extractio

Article · llama.cpp. hexagon: update to snapdragon-toolchain v0.8 and enable hexagon v85 ( #30854 ) hex-build: bump toolchain, enable v85 hex-build: bump toolchain to v0.8 in the CI and build scripts hex-build: add a little python script for swiv needed on some iot devices hex-build: fix flake8 warni

Article · llama.cpp. sycl: fix compiler warnings - missing prototypes, dead code, extra semicolons ( #30271 ) Website:

Article · llama.cpp. SearchInput already renders its own clear button. Hide the browser's native search cancel button to avoid showing both controls. Fixes #28627 Signed-off-by: aiqubits aiqubits@hotmail.com

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . Every release is signed with the same key introduced in commit 0112e53 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . Every release is signed with the same key introduced in commit 0112e53 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way

Article · llama.cpp. sycl : fix get_rows on reordered quantized weights ( #30430 ) Co-authored-by: Sam sam@garagellm.tail7dd858.ts.net Website:

Article · MarkTechPost. Nace AI released Drex Max, a multimodal decision model that reads text and screenshots and returns calibrated probabilities. Nace reports 67.59 on Decision Index 0.3.2 at $0.06 per 1M input tokens. The post Meet Drex Max: Nace AI’s Multimodal Decision Model for Computer-Use Agent

Article · llama.cpp. mtmd : match the transformers Qwen-VL resize ( #30736 ) Website:

Article · Hugging Face. Article from Hugging Face.

Article · llama.cpp. server: add --connect, allow p2p access to server outside LAN ( #28277 ) init wip wire it up make it clear that it's experimental add props.connect_code nits fix windows build server: do not advertise a connect code without the tunnel --connect-code on its own had no effect other

Article · llama.cpp. hybrid dna tokenizer fixes ( #30825 ) vocab : support Carbon-14B HybridDNA tokenizer Carbon-14B auto-tags untagged DNA runs and uses N-padded tail tokens (tail=pad_n). Also strip whitespace inside tags and handle a stray like the HF tokenizer. convert : fix special tokens and mer

Article · llama.cpp. opencl: add cooperative-k dp4a gemm for q4_0, q4_k and q6_k ( #30741 ) opencl: add cooperative-k dp4a GEMM for q4_0 opencl: add cooperative-k dp4a GEMM for q4_k opencl: add cooperative-k dp4a GEMM for q6_k opencl: fix alignment Website:

Article · llama.cpp. opencl: add get_rows for i32, cpy for i32 and f32 ( #30756 ) opencl: support i32 GET_ROWS opencl: support i32 and f32 copy conversions Website:

Article · llama.cpp. jinja : implement unique filter ( #30819 ) jinja : implement unique filter cont : inline key_of cont : fix alignment Website:

Article · Paperless-ngx. paperless-ngx 3.3.3 Bug Fixes Fix: fix global search input layout shifting @shamoon ( #14491 ) Fix: dont include doc content on global search results @shamoon ( #14488 ) Fix: remove some redundant pagers @shamoon ( #14487 ) Fix: prevent weird wrapping of dashboard widget column t

Article · llama.cpp. server : fix swa checkpoint invalidation logic ( #30666 ) server : fix swa checkpoint invalidation logic Assisted-by: Claude add testcase that fails with > n_past, but works with > n_prefix Website:

Article · noted.lol. Mitra brings events and tasks together in a self-hosted calendar, with Docker deployment and browser push reminders.

Article · llama.cpp. cann : refactor #ifdef blocks to avoid unreachable code after return ( #25273 ) Website:

Article · llama.cpp. CUDA: fix warning for nonexistent extra semicolon ( #30774 ) Website:

Article · llama.cpp. ggml-cpu: vectorize ggml_v_silu and ggml_v_expf ( #30791 ) Signed-off-by: Aaron Teo aaron.teo1@ibm.com Website:

Article · MarkTechPost. MiniMax has released M3.1 Flash, a 428B-parameter MoE model with about 23B active parameters and a 1M-token context. API pricing is $0.15/$0.60 per 1M tokens, with weights due October 23 The post MiniMax Just Launched M3.1 Flash: 428B MoE Agentic Model at $0.15/M Input appeared f

Article · llama.cpp. vendor : update cpp-httplib to 0.62.0 ( #30785 )

Article · llama.cpp. moe-cache : replace MoE LRU cache with a 50/50 SLRU cache ( #30637 )

Article · llama.cpp. qwen3 : bidirectional attention and dense_2 output projection ( #30607 ) qwen3 : bidirectional attention and dense_2 output projection Assisted-by: Grok convert : read use_bidirectional_attention in base Assisted-by: Grok Update conversion/base.py Co-authored-by: Sigbjørn Skjæret

Article · noted.lol. Home servers collect plenty of information about themselves. System logs record services restarting, login attempts, and other activity that usually goes unnoticed until something stops working. Finding anything meaningful can take longer than expected. I've always liked smaller

Article · llama.cpp. tools : reject truncated GGUF tensor data in split and export tools ( #30497 ) The problem here in the repo is gguf-split can silently produce corrupted tensor output when the input GGUF file is truncated. The fix is to add tensor-data bounds checks and short-read checks in gguf-

Article · llama.cpp. The fused moe_reduce kernel computes dst = sum(experts[e] * weights[e]) in a single dispatch, reading experts and weights while writing dst. The graph allocator computes tensor liveness on the unfused graph, where the inputs are fully consumed by earlier nodes of the fused group,

Article · llama.cpp. cuda: cap Stream-K blocks at the number of k-blocks Stream-K normally launches one CUDA block per SM. For small matrices, this can exceed the total number of output-tile/k-block pairs, leaving blocks with no work and adding launch and fixup overhead. Cap the grid at that count to

Article · n8n. 2.43.8 (2026-10-19) Bug Fixes core: Turn off expression spans by default ( #41310 ) ( e88162a )

Article · llama.cpp. CUDA: fix shared-memory WAR hazard in flash_attn_ext_vec ( #30573 ) CUDA: fix KQ write-after-read race in flash_attn_ext_vec Assisted-by: ChatGPT CUDA: apply review suggestions for KQ sync comments Co-authored-by: Johannes Gäßler johannesg@5d6.de Co-authored-by: Johannes Gäßler j

Article · llama.cpp. hexagon: FLASH_ATTN_EXT improvements for longer context and SWA models ( #30731 ) hex-fa: support for masked block skipping to reduce DMA fetching hex-fa: finetuning the hmx->hvx crossover thresholds (boosting gemma-4) hex-fa: update hmx gate to account for mdev.count Website:

Article · llama.cpp. hexagon: allow multistream FA ( #30218 ) Website:

Article · ComfyUI. Fix some crashing issues a few users experienced when running the MOSS image template workflow.

Article · ServeTheHome. In our GL.iNet Comet Q (GL-RMQ1) review, we see how this USB Type-C remote KVM device with Tailscale makes connectivity easy The post GL.iNet Comet Q GL-RMQ1 Review A Single USB Type-C Cable Remote KVM Device with Tailscale appeared first on ServeTheHome .

Article · ComfyUI. What's Changed [Partner Nodes] feat(Krea): add Krea 3 Image node by @bigcat88 in #17067 Update workflow templates to v0.11.85 by @comfyui-wiki in #17068 Use FLUX 3 Image in node display labels by @MaanavD in #17075 Better documentation about the comfy quant format. by @comfyanony

Article · noted.lol. Keep selected YouTube channels and playlists on your own server with Yarrtube, a lightweight Docker app built around yt-dlp.

Article · llama.cpp. cuda : fuse qwen4exp's SCALE -> SIGMOID -> SCALE -> hc_post chain ( #30602 ) Port of the Vulkan HC_POST_GATE fusion ( #29520 ). hc_combine builds the hc_post weights as 2 sigmoid(inject/hc); dsv4_hc_post_f32 now takes a gated template flag and computes scale_out sigmoid(scale_in*

Article · llama.cpp. ggml: support non-contiguous SUM (CPU, CUDA, Vulkan, WebGPU) ( #28851 ) ggml: support non-contiguous SUM (CPU, CUDA) tests: add non-contiguous SUM cases tests: SUM over non-contiguous views vulkan: support non-contiguous SUM webgpu: support non-contiguous SUM Website:

Article · noted.lol. iSponsorBlockTV connects to your television's YouTube app and automatically skips sponsor segments without replacing the app.

Article · llama.cpp. ggml-cuda: fattn do not check oob if k_VKQ_max = (k_VKQ_max - nbatch_fa) ( #30572 ) Website:

Article · vLLM. v0.32.0 Highlights This release features 814 commits from 363 contributors (121 new)! Structured decisions and generic parsers : the new POST /v1/systemone endpoint answers choice , noul and score questions over a structured state ( #59299 ), is served by default ( #60677 , #6069

Article · vLLM. [Doc][RL] Document the weight-transfer HTTP protocol for trainers wit…

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . Every release is signed with the same key introduced in commit 0112e53 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . Every release is signed with the same key introduced in commit 0112e53 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way

Article · llama.cpp. Restore ESIMD kernels on FFN path ( #30239 ) Reenable ESIMD on the FFN path Signed-off-by: Todd Malsbary todd.malsbary@intel.com Extend ESIMD fusion on FFN path to include Q8_0 Signed-off-by: Todd Malsbary todd.malsbary@intel.com Remove comment explicitly referencing ESIMD in tes

Article · llama.cpp. hexagon: add support for hadamard MUL_MAT and GET_ROW for flat and tiled Q1_0 and Q2_0 ( #30627 ) hex-hadamard: add support for hadamard ops hex-mm: reduce pkt counts in the q8_1 quantizer and vec_dot_q1 hex-q1/q2: add support for flat and tiled GET_ROWS for Q1_0 and Q2_0 Co-auth

Article · Ollama. Implement the Qwen3.5-backed decision scorer with noncausal attention, the publisher readout, image handling, and System One prompt encoding.

Article · llama.cpp. hexagon: add support for Q1_0 and Q2_0 ( #30122 ) hexagon: add support for q1_0 quant hexagon: enable Q1_0 weights in the 2D HMX MUL_MAT path hex-q1: cleanup gates, generalize tiling and remove in-loop divs hex-tiling: document tiled formats in developer.md and update htp-tensor.

Article · llama.cpp. ggml-cuda: HIP: use __builtin_amdgcn_perm for q1_0 unpacking ( #29927 ) Website:

Article · noted.lol. Homelable brings network discovery, live status checks and diagrams together to help document the machines behind a home server setup.

Article · LiteLLM. Verify Docker Image Signature All LiteLLM Docker images are signed with cosign . Every release is signed with the same key introduced in commit 0112e53 . Verify using the pinned commit hash (recommended): A commit hash is cryptographically immutable, so this is the strongest way

Article · ServeTheHome. This week NVIDIA launched their high-end RTX Spark N1X platform. Bringing their powerful GB10 SoC to Windows for the first time, the company is seeking to remake the Windows laptop market with their powerful, AI-focused SoC The post NVIDIA RTX Spark N1X Launch Overview: NVIDIA Ar

Article · llama.cpp. ui : Discover & Download Models from HF Hub ( #29584 ) common : resolve - download tags and list cached sidecars A Q4_0-mtp style tag now resolves the sidecar file when no model file matches it, so a solo draft or mmproj download actually pulls the file. Cached sidecar files list

Article · llama.cpp. CUDA: faster MMVQ for low-bit types ( #30517 ) CUDA: faster MMVQ for low-bit types ggml-cuda: hip: small nwarps GCN fix on Q3_K(ncols=2,3), Q6_K(ncols=4) Co-authored-by: thelittlefireman thelittlefireman3@gmail.com Website:

Article · Ollama. What's Changed win: prevent orphaned runners by @dhiltgen in #18500 gguf: add the q2_0 tensor and file types by @jmorganca in #18967 Launch clef in Ollama's cloud by @jberg5 in #19036 use identifiers for --ask mode question keys by @jberg5 in #19041 Full Changelog : v0.40.5...v0.

Article · llama.cpp. tests: give the random model generator a per-tensor filler table ( #30674 ) tests: give the random model generator a per-tensor filler table Muse Spark makes test-llama-context fail on state_load, cp_h and cp_d on about six percent of the random seeds, with an NMSE of 1.0e-05 to

Article · llama.cpp. cuda : fix -Wtautological-compare in fattn vec dispatch ( #30519 ) Signed-off-by: Adrien Gallouët angt@huggingface.co Website:

Article · noted.lol. Your homelab may be costing more to run than it needs to. Here's why separating essential services from your lab can save electricity.

Article · llama.cpp. cuda : add grouped experts top-k fusion ( #29181 ) fuse grouped experts add tests hip fix add to graph_optimize apply suggestion from am17an hip fix fix hip fix fix fix fix gate on hip version instead Website:

Article · llama.cpp. model : add Kolibri-1 (Aleph-Alpha) support ( #30673 ) model : add Kolibri-1 (Aleph-Alpha) support Kolibri-1 is a bilingual German/English MoE reasoning model with 78B parameters and 3.46B active per token: 50 layers, 384 routed experts plus one shared expert. Attention interleav

Article · llama.cpp. model : Granite switch decision 256 options ( #30693 ) feat(convert): Attempt to come up with 256 single-token, semantic option markers Each switch model needs to identify a collection of single-token option markers that can be mapped to their respective logits for the output pro

Article · ComfyUI. Article from ComfyUI.

Article · llama.cpp. sycl: fix prompt-processing regression after Q4_0/Q8_0 weight reorder ( #30374 ) Website:

Article · llama.cpp. hexagon: vectorize softplus with HVX ( #30435 ) softplus_f32 and tile_softplus_f32 ran a scalar logf(1.0f + expf(x)) loop. Replace both with a shared hvx_softplus_f32_aa() helper built from the backend's existing hvx_vec_exp_f32 / hvx_vec_log_f32 primitives, so no new numerical a

Article · vLLM. Signed-off-by: Vadim Gimpelson vadim.gimpelson@gmail.com Co-authored-by: Claude noreply@anthropic.com

Article · llama.cpp. opencl: Add MxFP4 MoE bin kernel for Adreno X1 devices ( #30691 ) opencl: add bin kernel gemm_moe_mxfp4_q8_1_dp4a_ila_ldck for X1 opencl: fix whitespace opencl: undo unnecessary change opencl: enable dp4a clc for X1E Co-authored-by: Li He lih@qti.qualcomm.com Website:

Article · MarkTechPost. Perplexity has open-sourced SPACE, the sandbox platform behind every task in Perplexity Computer. The Perplexity SPACE sandbox gives each agent task its own Firecracker microVM. Credentials and outbound traffic are controlled from outside that VM. The code ships under Apache 2.0

Article · llama.cpp. model: add Beam (Reflection) ( #30670 ) Website:

Article · Ollama. Behind a proxy, defer redirect-target DNS to the proxy; without one, keep the resolve-and-pin path. Trust default-registry/CDN hostnames with no local DNS. Fix a nil CancelFunc crash in blobDownload.release.

Article · noted.lol. FileBrowser Quantum 2.0 introduces activity logs and improved file permissions, alongside a SQLite migration and important security fixes.

Article · llama.cpp. model : Granite Switch Decision ( #30676 ) feat(convert): Add decision.chat to support decision models that can also chat Branch: GraniteSwitchDecision AI-usage: full (Claude + Sonnet 5.5) Signed-off-by: Gabe Goodhart ghart@us.ibm.com feat(convert): Add conversion for Granite Swi

Article · llama.cpp. common: honor HTTP_PROXY/HTTPS_PROXY env vars in http client ( #21752 ) common: honor HTTP_PROXY/HTTPS_PROXY env vars in http client common_http_client() was ignoring standard proxy env vars, causing downloads via -hf and similar flags to fail silently in environments behind an H

Video · Techno Tim. Did you know Linux can cache NFS files on a local SSD? I somehow missed this feature, so I decided to try it out. In this video, I'll show you how to set up FS-Cache, CacheFiles, and cachefilesd on Ubuntu, how caching and eviction work, and how much faster repeated reads can be.

Article · ServeTheHome. We review the Microsoft Surface Ultra to see how the NVIDIA RTX Spark powered laptop fares and try to answer who the platform is for The post Microsoft Surface Laptop Ultra 128GB Review appeared first on ServeTheHome .

Article · ServeTheHome. We review the Microsoft Surface Ultra to see how the NVIDIA RTX Spark powered laptop fares and try to answer who the platform is for The post Microsoft Surface Laptop Ultra Review appeared first on ServeTheHome .

Article · noted.lol. Libreboot 26.10 gives older computers a second chance with modern UEFI support, new hardware compatibility, and expanded Chromebook builds.

Article · selfh.st. You either die open source, or live long enough to see yourself become open core

Article · Open WebUI. Added 🔐 End-to-end encrypted chats. Direct messages and private group channels can now hold a temporary end-to-end encrypted chat: a member picks Encrypted Chat in the channel menu, the other members see "Encrypted chat in progress" with a Join link, messages of up to 16 KiB are

Article · llama.cpp. hexagon: correct HMX FA sinks ( #30581 ) hexagon: correct HMX FA sinks hex-fa: vectorize sinks in hmx kernels Co-authored-by: Max Krasnyansky maxk@qti.qualcomm.com Website:

Article · llama.cpp. model : MTP support for Mellum2.1 ( #30612 ) model : MTP support for Mellum model : inline rope construction in Mellum Website:

Article · noted.lol. Learn how to serve your music with spotify-like features, but all locally.

Article · llama.cpp. [SYCL]: fix memory errors in mul_mat, split buffer, host pool ( #29889 ) Website:

Article · llama.cpp. common : read ngram cache parts by reference instead of copying ( #30061 ) Website:

Article · llama.cpp. cpu : reduce AVX2 tiled accumulators together ( #30296 ) Assisted-by: Codex Website:

Article · llama.cpp. ggml-cpu: refactor s309x repack helper functions ( #30588 ) Signed-off-by: Aaron Teo aaron.teo1@ibm.com Website:

Article · llama.cpp. sycl: wide stores for the q4_K/q5_K weight dequant ( #29696 ) Website:

Article · llama.cpp. tests : catch vanished rows in the fusion baseline ( #30618 ) tests : update the Metal fusion baseline Re-record the MTL fusion baseline with the current build: 276 rows over 131 dummy models. Assisted-by: pi:llama.cpp/ggml-org/Qwen3.8-Flash-Next-GGUF (MacBook, M5 Max, 128GB) tes