Two findings from benchmarking 7 models across 5 agent frameworks:
1. Qwen family gets 100% tool calling across every framework tested. Non-Qwen models (Llama, DeepSeek-R1) vary wildly \u2014 40% to 100% depending on framework.
2. smolagents (HuggingFace) sidesteps structured function calling entirely by using code generation. DeepSeek-R1 goes from 40% with structured FC to 100% with smolagents.
Speed-wise, MLX's unified memory means zero CPU\u2194GPU copies. On an M3 Ultra: Qwen3.5-9B hits 108 tok/s (vs ~41 on Ollama), Qwen 3.6 35B does 100 tok/s with only 3B active params.
The full benchmark data is in the README. Happy to discuss the MLX performance characteristics or tool calling architecture.","title":null,"type":"comment","url":null},{"author":"c0rruptbytes","children":[{"author":"taylorhou","children":[{"author":"Johnny_Bonk","children":[],"created_at":"2026-05-06T16:42:39.000Z","created_at_i":1778085759,"id":48038324,"options":[],"parent_id":48027023,"points":null,"story_id":47816238,"text":"This looks somewhat interesting. Is the premise that if you have a strong Mac, you can rent out your hardware for someone to run inference on it?","title":null,"type":"comment","url":null}],"created_at":"2026-05-05T19:05:22.000Z","created_at_i":1778007922,"id":48027023,"options":[],"parent_id":47974394,"points":null,"story_id":47816238,"text":"is this available for other open source projects? i'm stealing tokens from my employer effectively and keep hitting my limits re: codex tokens o_O
i'm working on github.com/teale-ai (distributed inference)","title":null,"type":"comment","url":null}],"created_at":"2026-05-01T13:09:54.000Z","created_at_i":1777640994,"id":47974394,"options":[],"parent_id":47816238,"points":null,"story_id":47816238,"text":"i use this, big fan - i have unlimited codex tokens if you ever need some dev assistance
Would definitely love benchmarks against omlx and fast-mlx one day (i also have a 256gb m3 ultra)","title":null,"type":"comment","url":null}],"created_at":"2026-04-18T14:33:55.000Z","created_at_i":1776522835,"id":47816238,"options":[],"parent_id":null,"points":9,"story_id":47816238,"text":null,"title":"Show HN: Rapid-MLX \u2013 Run local LLMs on Mac, 2-3x faster than alternatives","type":"story","url":"https://github.com/raullenchai/Rapid-MLX"}