Coding Tools
llama.cpp v0.9.0 adds FP8 and audio generation
llama.cpp Releaseshigh confidence
Why it matters
- →Native FP8 E4M3 support enables efficient inference on hardware that requires this standard without external conversion steps.
- →Unified KV cache optimizations reduce memory overhead and improve throughput for batched multi-sequence requests.
- →Expanded audio capabilities allow the runtime to handle TTS and transcription, broadening its use beyond text generation.
llama.cpp has released version 0.9.0, introducing native support for OCP FP8 E4M3 quantization and optimized unified KV caching for multi-sequence inference. The update expands multimodal capabilities by adding Chatterbox TTS for audio generation and Nemotron 3.5 Transcribe for speech-to-text, alongside support for new model architectures like Metis-1.6 and Mistral Large 4. Under the hood, ggml v0.29.0 brings NCCL-based multi-node tensor parallelism and performance improvements across CPU, CUDA, and Metal backends. The release also includes server refinements and UI updates to better handle tool calls and persistent storage.
Read originalThe story around this
Earlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
How we got here
- 1Llama.cpp b10326 Release Enhances Audio Processing
llama.cpp Releases · August 8, 2026 · Same story
- 2llama.cpp adds Soprano TTS support
llama.cpp Releases · October 24, 2026 · Same story
Follow this story
Open the full story →Liquid AI releases LFM3 multimodal models
4 developments
- Oct 27 · Hugging Face Blog
Liquid AI releases LFM3 multimodal models - Oct 28 · llama.cpp Releases
llama.cpp adds LiquidAI LFM3 model support↳ llama.cpp adds support for LiquidAI LFM3 models to enable local inference on Apple Silicon and AMD ROCm GPUs. - Oct 30 · llama.cpp Releases
llama.cpp adds OCP FP8 E4M3 support↳ llama.cpp adds OCP FP8 E4M3 quantization support with fast CPU kernels and fixes for fused QKV projections. - Nov 5 · llama.cpp Releases
llama.cpp v0.9.0 adds FP8 and audio generation (This article)↳ llama.cpp v0.9.0 adds native OCP FP8 E4M3 support, unified KV cache optimization, and audio generation via Chatterbox TTS and Nemotron 3.5
More from llama.cpp Releases
Coding Toolscoding
llama.cpp b12173 adds OpenCL repeat ops
This release quietly patches a gap in llama.cpp's OpenCL backend by adding support for the 'repeat' operation across f16, i32, i16, and bf16 types. For users relying on AMD or Intel GPUs via OpenCL rather than Vulkan or ROCm, this means fewer fallbacks to slower CPU execution for models that use these tensor operations. It’s a targeted fix that improves stability and performance on non-NVIDIA hardware without changing the core inference engine. The update also includes standard binaries for CUDA 13 and ROCm 10.0, keeping the stack current with latest driver ecosystems.
llama.cpp Releases
Coding Toolscoding
llama.cpp b12174 adds ROCm 10 and CUDA 13
This release quietly closes the hardware gap for non-NVIDIA users by adding default support for ROCm 10.0 on both Linux and Windows, allowing AMD GPU owners to run local models without complex configuration. It also brings CUDA 13.5 builds to Ubuntu and Windows, future-proofing the toolchain for newer NVIDIA architectures while maintaining backward compatibility with CUDA 12. Apple Silicon KleidiAI is disabled in this build, a notable regression for Mac users who rely on that specific optimization. The update solidifies llama.cpp's position as the most versatile inference runtime available, covering everything from Snapdragon NPUs to Intel Arc via SYCL.
llama.cpp Releases
Coding Toolscoding
llama.cpp b12175 fixes CI hangs, adds ROCm 10 and CUDA 13
This release stabilizes the build pipeline by fixing a threading bug that caused test hangs on self-hosted runners, ensuring more reliable binaries for everyone. More importantly, it brings first-class support for ROCm 10.0 and CUDA 13 across Linux and Windows, closing the gap with NVIDIA's latest driver stacks. AMD users finally get parity with CUDA builds without manual configuration tweaks. The update also expands Apple Silicon coverage to include KleidiAI, though it remains disabled by default in this build. It’s a maintenance-heavy release that quietly modernizes the backend support matrix.
llama.cpp Releases
More in Coding Tools
Coding Toolscoding
Claude Code v2.1.319: MCP backgrounding and gateway controls
This release finally brings proper asynchronous handling to Model Context Protocol tasks, allowing Claude Code to offload long-running tool calls to the background instead of blocking the main thread. For enterprise users, the new pooled budget rules and JSON logging make the Claude apps gateway significantly more manageable for large teams. The update also tightens error reporting in auto mode, ensuring developers get clear reasons when actions are denied rather than silent failures. It’s a pragmatic patch that improves reliability for complex agent workflows.
Claude Code Releases
© GitHub ChangelogCoding Toolscoding
GitHub Copilot Code Review API and UI updates
GitHub is finally exposing Copilot code review configuration to the REST and GraphQL APIs, closing a major gap for enterprise automation. Administrators can now programmatically control review effort, enablement, and custom instructions across organizations and repositories, moving beyond manual web interface tweaks. The update also refines the user experience with compact overviews, explicit attribution of AI sources like MCP servers, and better handling of bot-authored PRs. This shift makes Copilot Review a first-class citizen in CI/CD pipelines rather than just a UI feature.
GitHub Changelog
© GitHub ChangelogCoding Toolscoding
GitHub Copilot Dashboard Adds Chat and Agent Metrics
GitHub finally stops treating chat as a passive afterthought in its enterprise analytics. By adding Copilot Chat and the VS Code Agents window to the impact dashboard, the platform now tracks nine distinct experiences instead of just seven. The rebranding of 'Passive users' to 'Chat-first users' is a semantic correction that aligns metrics with actual developer behavior rather than arbitrary adoption thresholds. This gives engineering leaders a complete picture of how AI is actually being used across their codebases, closing the visibility gap between typing commands and writing code.
GitHub Changelog