6 ms·
Llama.cpp now supports tool calling (OpenAI-compatible)
- ochafik 2y agoSupports 8+ models’ native formats (Llama, Functionary, Hermes, Mistral, Firefunction, DeepSeek) + generic fallback for all other models. Runs anywhere; on a Mac: brew install llama.cpp llama-server --jinja -fa -hf bartowski/Qwen2.5-7B-Instruct-GGUF:Q4_K_M Still fresh / lots of bugs to discover, feedback welcome!