AnalysisAI ModelsAugust 21, 2026

Ornith-1.5-35B-A3B-NInfer hits 250 tok/s on 5090

A Reddit user reports running Ornith-1.5-35B-A3B-NInfer on Ninfer for Windows, calling it the best local model for interactive use with phenomenal speed. It achieves 250 tok/s with 5-8k prefill on a 5090.

1 source

More stories today

Open the live feed