AnalysisAI ModelsSeptember 7, 2026

Qwen3.8-Flash-Next-oQ4e-mtp hits 47 tok/s on Apple Silicon

Across 22 community benchmark runs, Qwen3.8-Flash-Next-oQ4e-mtp peaks at 47.0 tok/s and averages 40.7 tok/s, with the fastest results on an M4 Max. It needs at least 68.9 GB of memory and scores 72.6/100 for coding and 83.5/100 for agentic tasks.

1 source

More stories today

Open the live feed