

llama.cpp
#2 v Lokální LLM runtimeGgml Org · od 13. November 2026 (v1.0.0) · 7× · naposledy 15. 11. 2026
88
Momentum
llama.cpp is an open-source C/C++ inference library and server for running LLMs on consumer hardware and in the cloud. Version 1.0.0 (libllama) and ggml 1.0.0 were released on November 13, 2026 and now follow semantic versioning. The official product site is llama.app; the code is hosted at github.com/ggml-org/llama.cpp.
Vývoj momenta
18.08.16.11.
Vlastnosti
| Throughput/Latency | CPU+GPU hybrid inference for models larger than VRAM; v1.0.0: host-memory MoE expert cache with configurable CPU share |
| License | MIT |
| Platform | Windows, macOS, Linux, Docker, Homebrew, Nix, MacPorts |
| Price | Free (open source) |
| Protocol Compatibility | OpenAI-compatible (/v1/chat/completions, /v1/embeddings, /v1/models) and Anthropic-compatible (/v1/messages) |
| Release Date | Nov 13, 2026 (v1.0.0, libllama and ggml) |
| Supported Models/Providers | More than 150 model architectures (v1.0.0 adds Kimi-K3.1-Pro, Ripple, xing4_0) |