Same model. Same Mac.

2.4×

speed difference.

Bench your model!

asiai bench

$ asiai bench -m qwen3.5

lmstudio 72.6 tok/s

ollama 30.4 tok/s

Winner: lmstudio (2.4× faster)

7 engines GPU observability Open Source Zero dependencies Leaderboard MCP Claw-ready Agent-ready