Comment on Show HN: Running 104GB Qwen3.8-Flash-Next on 48GB Mac with at ~12 tok/sComments−securecloudgrou9dWould love to see metrics of model performance, comparison with oMLX/OLLAMA/others. Any tooling for local optimization on hardware.−securecloudgrou9dTo be more clear, I see you have slotstream doctor --sim-ram N, but extended tooling and optimization for exact local hardware would add value (MTPLX has a nice interface for example).
Comments
Would love to see metrics of model performance, comparison with oMLX/OLLAMA/others. Any tooling for local optimization on hardware.
To be more clear, I see you have slotstream doctor --sim-ram N, but extended tooling and optimization for exact local hardware would add value (MTPLX has a nice interface for example).