AnalysisAI ModelsAugust 9, 2026

LFM 2.6B reaches 260T/s in user's local tests

Reddit user reports 260T/s tg and 20k pp on 3090s, calling the phone-sized model surprisingly good for quick tasks like scanning long text or summarizing.

1 source

Daily brief

Get tomorrow's AI brief in your inbox

More stories today

Open the live feed
LFM 2.6B reaches 260T/s in user's local tests — AIBriefs