raullen

Built this to run coding agents locally on Apple Silicon. The main problem I kept hitting: most models fail at structured tool calling, and existing servers are slow on MLX.

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 — 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↔GPU 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.

I've been working on a fork of vllm-mlx (OpenAI-compatible LLM server for Apple Silicon) to make it actually usable for coding agents. The upstream project is great but was missing production-grade tool calling, reasoning separation, and multi-turn performance.

  What I added (37 commits):

  - Tool calling that works — streaming + non-streaming, supports MiniMax and Hermes/Qwen3 formats. 4/4 accuracy on structured function calling benchmarks.
  - Reasoning separation — MiniMax-M2.5 mixes reasoning into its output with no tags. Built a heuristic parser that cleanly separates reasoning from content (0% leak rate, was 60%
   with the generic parser).
  - Prompt cache for SimpleEngine — persistent KV cache across requests. On 33K-token coding agent contexts: TTFT goes from 28s to 0.3s on cache hit. This is the single biggest
  improvement for multi-turn use.
  - 1500+ tests — parsers, engine, server, tool calling. The upstream had minimal test coverage.

  Benchmarks (Mac Studio M3 Ultra, 256GB):

  Qwen3-Coder-Next-6bit (80B MoE, 3B active):
  - Decode: 65 tok/s
  - Prefill: 1090-1440 tok/s
  - TTFT (cache hit, 33K context): 0.3s

  MiniMax-M2.5-4bit (229B MoE):
  - Decode: 33-38 tok/s
  - Deep reasoning with tool calling

  I built this to run OpenClaw locally on my Mac instead of paying for cloud APIs. Qwen3-Coder-Next at 65 tok/s with tool calling is genuinely usable — not a toy demo.

  Quick start:

  pip install git+https://github.com/raullenchai/vllm-mlx.git
  python -m vllm_mlx.server \
    --model lmstudio-community/Qwen3-Coder-Next-MLX-6bit \
    --tool-call-parser hermes --port 8000

  GitHub: https://github.com/raullenchai/vllm-mlx

Running a long Claude Code session? Need to step away from your desk? Claw lets you monitor and control Claude Code from any device with a browser.

See what Claude is doing in real-time from any screen Send quick responses (yes/no/continue) with one tap Interrupt with Ctrl+C when things go sideways Monitor everything — sessions, windows, panes, git status, system stats

Mine dies for 10min