Show HN: llamafile 0.10.0 rebuilt, Qwen3.5, lfm2, Anthropic API
7 hours ago
- #AI-models
- #llamafile
- #portability
- llamafile 0.10.0 unifies portability and modern model features.
- Includes bundle weights, multimodal models, tool calling, and Anthropic Messages API support.
- Rebuilt from the ground up for easier updates with upstream dependencies.
- Supports Qwen3.5 models for vision, lfm2 for tool calling, and Claude code via Anthropic Messages API.
- Features APE executables, full llama.cpp server features, multimodal terminal chat, multiple UIs, GPU support (Metal/CUDA), and CPU optimizations.
- Pre-built llamafiles available with models ranging from 0.6B to 27B parameters.
- Future plans include feature parity with older versions, easier bundling, Vulkan support, and issue fixes.
- Older llamafile versions and models still accessible for users needing previous features.