https://freedom.tech/project/llama-cpp/ Every llama.cpp release and update tracked by Freedom.Tech. en Freedom.Tech is a Foundation project. https://foundation.xyz Freedom.Tech, a Foundation project (https://foundation.xyz) https://freedom.tech/posts/2026-08-27-llama-cpp/ https://freedom.tech/posts/2026-08-27-llama-cpp/ Thu, 27 Aug 2026 05:22:38 GMT Adds nanbeige4.2-3B model support https://freedom.tech/posts/2026-08-26-llama-cpp/ https://freedom.tech/posts/2026-08-26-llama-cpp/ Wed, 26 Aug 2026 12:44:43 GMT Ui: disable the npm UI build (LLAMA BUILD UI=OFF) https://freedom.tech/posts/2026-08-25-llama-cpp/ https://freedom.tech/posts/2026-08-25-llama-cpp/ Tue, 25 Aug 2026 17:26:21 GMT Scopes Qwen3-coder chat workarounds https://freedom.tech/posts/2026-08-25-llama-cpp-0-3-0/ https://freedom.tech/posts/2026-08-25-llama-cpp-0-3-0/ Tue, 25 Aug 2026 10:22:58 GMT Llama.cpp 0.3.0 adds dots3-note multimodal support, tensor-split for DeepSeek 4, multi-token prediction for GLM-4.5-Air, and WebP image decoding https://freedom.tech/posts/2026-08-24-llama-cpp/ https://freedom.tech/posts/2026-08-24-llama-cpp/ Mon, 24 Aug 2026 13:16:46 GMT Metal flash-attention tuning: 53 new f16 instantiations, per-device dispatch tables https://freedom.tech/posts/2026-08-23-llama-cpp/ https://freedom.tech/posts/2026-08-23-llama-cpp/ Sun, 23 Aug 2026 18:45:59 GMT GLM-4.5-Air MTP support added to llama.cpp for improved model inference efficiency https://freedom.tech/posts/2026-08-22-llama-cpp/ https://freedom.tech/posts/2026-08-22-llama-cpp/ Sat, 22 Aug 2026 23:40:48 GMT Fixes clang LTO build issue https://freedom.tech/posts/2026-08-21-llama-cpp/ https://freedom.tech/posts/2026-08-21-llama-cpp/ Fri, 21 Aug 2026 22:56:01 GMT CI switches to LLVM OpenMP on Windows; Drops non-redist debug build https://freedom.tech/posts/2026-08-21-llama-cpp-0-2-0/ https://freedom.tech/posts/2026-08-21-llama-cpp-0-2-0/ Fri, 21 Aug 2026 18:32:48 GMT Llama.cpp 0.2.0 syncs ggml to 0.21.0, adds backend optimizations for CUDA/Metal/SYCL/OpenCL, and improves quantization kernel performance across hardware https://freedom.tech/posts/2026-08-20-llama-cpp/ https://freedom.tech/posts/2026-08-20-llama-cpp/ Thu, 20 Aug 2026 02:36:34 GMT Adds dedup-cache-models preset option to server https://freedom.tech/posts/2026-08-18-llama-cpp/ https://freedom.tech/posts/2026-08-18-llama-cpp/ Tue, 18 Aug 2026 10:43:36 GMT Fixes LFM2 image tiling threshold; Cross-platform refactor https://freedom.tech/posts/2026-08-18-llama-cpp-0-1-2/ https://freedom.tech/posts/2026-08-18-llama-cpp-0-1-2/ Tue, 18 Aug 2026 10:23:10 GMT Llama.cpp synced GGML 0.20.2, added integer tokenizer scores, improved CUDA performance on DGX, and fixed xcframework builds https://freedom.tech/posts/2026-08-17-llama-cpp/ https://freedom.tech/posts/2026-08-17-llama-cpp/ Mon, 17 Aug 2026 13:59:43 GMT CI; Explicitly pushes release tag before creating GitHub release https://freedom.tech/posts/2026-08-16-llama-cpp/ https://freedom.tech/posts/2026-08-16-llama-cpp/ Sun, 16 Aug 2026 11:21:52 GMT Removes some ggml concat calls; Model-layer refactor https://freedom.tech/posts/2026-08-15-llama-cpp/ https://freedom.tech/posts/2026-08-15-llama-cpp/ Sat, 15 Aug 2026 20:14:19 GMT Adds Kimi-K3 text-model support with hybrid KDA and MLA attention https://freedom.tech/posts/2026-08-14-llama-cpp/ https://freedom.tech/posts/2026-08-14-llama-cpp/ Fri, 14 Aug 2026 19:22:14 GMT Llama.cpp now passes reasoning_effort parameter through chat templates, enabling fine-grained control over model reasoning depth in inference https://freedom.tech/posts/2026-08-13-llama-cpp/ https://freedom.tech/posts/2026-08-13-llama-cpp/ Thu, 13 Aug 2026 22:11:34 GMT OpenVINO backend gains Qwen3.5 support, memory optimizations for GPU inference, and fixes for stateful RoPE accuracy and recurrent state handling https://freedom.tech/posts/2026-08-12-llama-cpp/ https://freedom.tech/posts/2026-08-12-llama-cpp/ Wed, 12 Aug 2026 04:52:57 GMT Build b10375 tightens bare function parsing for Qwen models https://freedom.tech/posts/2026-08-11-llama-cpp/ https://freedom.tech/posts/2026-08-11-llama-cpp/ Tue, 11 Aug 2026 10:48:48 GMT Switches ROCm target to 7.14, the first production release using TheRock build system https://freedom.tech/posts/2026-08-10-llama-cpp/ https://freedom.tech/posts/2026-08-10-llama-cpp/ Mon, 10 Aug 2026 16:24:12 GMT Added MTP (Multi-Token Prediction) support for Nemotron models, enabling faster inference for this architecture https://freedom.tech/posts/2026-08-08-llama-cpp/ https://freedom.tech/posts/2026-08-08-llama-cpp/ Sat, 08 Aug 2026 16:00:17 GMT Corrects server info endpoint to report the isolate working directory when a tools runtime is configured https://freedom.tech/posts/2026-08-07-llama-cpp/ https://freedom.tech/posts/2026-08-07-llama-cpp/ Fri, 07 Aug 2026 21:23:15 GMT TTS timing now accounts for full vocoder pass instead of single trailing window https://freedom.tech/posts/2026-08-06-llama-cpp/ https://freedom.tech/posts/2026-08-06-llama-cpp/ Thu, 06 Aug 2026 01:33:36 GMT AMD ROCm CI support for gfx1151; No inference changes https://freedom.tech/posts/2026-08-05-llama-cpp/ https://freedom.tech/posts/2026-08-05-llama-cpp/ Wed, 05 Aug 2026 20:02:56 GMT Fixed directory traversal vulnerabilities in file search, improved path handling on Windows with UTF-8 conversion, and added cache expiration for UI picker searches https://freedom.tech/posts/2026-08-04-llama-cpp/ https://freedom.tech/posts/2026-08-04-llama-cpp/ Tue, 04 Aug 2026 18:56:20 GMT Adds a per-conversation working directory with file picker; Agents can now treat path-like queries as directory navigation https://freedom.tech/posts/2026-08-03-llama-cpp/ https://freedom.tech/posts/2026-08-03-llama-cpp/ Mon, 03 Aug 2026 05:11:26 GMT Implements DSv4 Lightning Indexer and f16 Lightning Indexer for Metal; Affects 128-dimensional 64-head inputs on Apple silicon https://freedom.tech/posts/2026-08-02-llama-cpp/ https://freedom.tech/posts/2026-08-02-llama-cpp/ Sun, 02 Aug 2026 18:57:43 GMT WebGPU backend adds f16 repeat support; MacOS KleidiAI build disabled https://freedom.tech/posts/2026-08-01-llama-cpp/ https://freedom.tech/posts/2026-08-01-llama-cpp/ Sat, 01 Aug 2026 16:46:07 GMT Llama.cpp now preserves reasoning content in chat history, enabling multi-turn conversations to leverage previous model reasoning with --reasoning-preserve flag https://freedom.tech/posts/2026-07-31-llama-cpp/ https://freedom.tech/posts/2026-07-31-llama-cpp/ Fri, 31 Jul 2026 19:25:45 GMT Adds driver-version check for Intel GPUs on Windows to avoid crashes; No action unless running that config https://freedom.tech/posts/2026-07-30-llama-cpp/ https://freedom.tech/posts/2026-07-30-llama-cpp/ Thu, 30 Jul 2026 17:21:07 GMT Tests alternative convolution layout; Extends layout checks for conv2d kernel