Anthropic has released Claude Code version 2.1.321, introducing Claude Fable 5.2 as the new default model with a 1 million token context window. The update includes critical security fixes for MCP OAuth requests and Bash sandbox permission checks to prevent token leakage and unauthorized file access. Additional patches address stability issues in Remote Control, background agents, and network connectivity, ensuring more reliable performance during long coding sessions. This release prioritizes security hardening and workflow consistency over new features.
Read originalEarlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
Claude Code Releases · September 2, 2026 · Same story
Claude Code Releases · October 4, 2026 · Same story
GitHub Changelog · November 9, 2026 · Related
This release quietly patches a critical memory overflow in the mtmd module that caused crashes when loading HunyuanVL models. The issue stemmed from F64 position embeddings overflowing an F32 buffer, a subtle bug that broke multimodal inference for users relying on this specific architecture. By enforcing strict F32 requirements, llama.cpp stabilizes support for Tencent's vision-language model without adding new features or changing the core API. It’s a necessary maintenance update that keeps local multimodal workflows from breaking unexpectedly.
This release quietly extends llama.cpp’s hardware coverage with ROCm 10.0 binaries for both Linux and Windows, finally bringing AMD RDNA/CDNA users into the fold without manual compilation. The standout feature is native support for Qwen3.5 models via tensor merging during conversion, a necessary step as this new architecture gains traction in the open-source community. By shipping pre-built CUDA 13 binaries alongside the older 12.x versions, the project ensures forward compatibility with newer NVIDIA GPUs out of the box. There are no new quantization algorithms or performance breakthroughs here; it is purely an infrastructure update to keep pace with the rapidly shifting landscape of model architectures and GPU drivers.
This release quietly fixes a critical reliability gap in llama.cpp's server mode: when a client disconnects or times out, pending Model Context Protocol (MCP) tool calls are now properly cancelled. This prevents orphaned processes and resource leaks during long-running agent workflows. The build matrix also expands with ROCm 10.0 support for both Linux and Windows, alongside CUDA 13.5 binaries, keeping the project aligned with the latest GPU driver ecosystems. While KleidiAI on Apple Silicon is temporarily disabled, the core inference engine remains robust for local AI operations.