llama.cpp Releases

A patch release that fixes a critical issue where static-frame deduplication was freezing `onUpdate`-driven motion sequences.

Sep 9, 2026

A major release focusing on render stability, improved lint coverage, and enhanced skill reliability across the platform.

Sep 8, 2026

Performance improvements to the Studio preview server, reducing memory usage and latency by serving byte ranges instead of reading whole files.

Sep 7, 2026

Enhanced file handling capabilities with site capture improvements and performance optimizations across core, engine, studio and producer components.

Sep 6, 2026

Critical stability fix preventing Studio from freezing when external tools write project files in rapid bursts.

Sep 5, 2026

llama.cpp 0.4.0 introduces support for Qwen3.8-Flash-Next and Nemotron-3-Puzzle models, adds lazy tensor reading and per-slot server context limits, and updates ggml to 0.23.0 with sparse flash attention and RDMA capabilities.

Sep 4, 2026

Open Recorder release focusing on geometry fixes for area selection, CSS isolation, and user interaction improvements like canceling countdowns.

Aug 31, 2026

Open Recorder release fixing HUD visibility issues and adding new background media options including auto-zoom and 12 new wallpapers.

Aug 27, 2026

Open Recorder release ensuring capture overlays remain visible above fullscreen applications and updating documentation.

Aug 27, 2026

Open Recorder release addressing camera session race conditions and HUD visibility issues, alongside documentation improvements.

Aug 26, 2026

Open Recorder release introducing fullscreen HUD support, multi-monitor area dragging, and new camera source options (Continuity/Desk View).

Aug 26, 2026

Major version bump adding dots3-note multimodal model support, GLM-4.5-Air MTP support, and ggml v0.22.0 updates including meta-backend tensor split and Metal parallel compilation.

Aug 25, 2026

Major version bump introducing KleidiAI SME2 support, SYCL kernel updates, UI refactor, and deep support for DeepSeek 4 models.

Aug 21, 2026

Enables tensor split support for LFM2 and LFM2MOE models in Tensor Parallelism (TP) mode.

Aug 21, 2026

Fixes a critical bug in the Metal Tensor API mat-mat kernel where reading out-of-bounds memory caused NaNs when the K dimension was not a multiple of 32.

Aug 21, 2026

Adds a command-line argument `--mmproj-device` to the MTMD (Multi-Modal Text-to-Model) backend to allow specifying the device for the multimodal projection model.

Aug 21, 2026

Fixes a potential overflow in the Vulkan Flash Attention (FA) MMQ kernel where Q quantization calculations used denormalized floats.

Aug 21, 2026

Optimizes the macOS Metal backend by dequantizing the Key-Value (KV) cache only for large batches to reduce overhead and improve efficiency.

Aug 21, 2026

Updates the CI build process to use LLVM's OpenMP library instead of the Microsoft debug version, improving compatibility and packaging.

Aug 21, 2026

Adds a lazy-loading feature to the server router that delays the loading of startup models until after the main server setup is complete.

Aug 21, 2026