‹ BackHN Continuity

Thread

Run Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/s

741 points · 330 comments · snehesht

  1. slashtom · · focus · HN ↗
    Qwen 3.8 Flash Next with mtplx on 128GB M5 Max is amazing, full context I'm getting around 72tok/s, it has replaced the frontier models for me.
Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.