Engines / antonellof / frink

frink
Read this first. Apple Metal only: the llama.cpp baseline build is unknown, and frink is much slower than llama.cpp on CPU and CUDA
Loads GGUF models in pure Rust, with quantized Metal kernels and lossless speculative decoding.
Benchmarks
- On Apple M2 Pro (10c/6p), macOS 26.6.2, Metal with Gemma-3-1B-IT (Q8_0), it decodes at 122 tok/s, against 83.2 tok/s for llama.cpp (+47%). Source
- On Apple M2 Pro (10c/6p), macOS 26.6.2, Metal with Llama-3.2-1B-Instruct (Q4_K_M), it decodes at 157 tok/s, against 149 tok/s for llama.cpp (+5%). Source
- On Apple M2 Pro (10c/6p), macOS 26.6.2, Metal with Qwen3-0.6B (Q8_0), it decodes at 190 tok/s, against 116 tok/s for llama.cpp (+64%). Source
Listing
Listed: Passed all seven checks.
Security review
Not reviewed yet.
Install
To install it, follow the project's own instructions on GitHub. The engine runs with your user's permissions. macOS doesn't sandbox it, so install only engines you trust.
Something wrong or out of date? Write to help@kalsa.io, or ask us to remove this page.