LiquidAI LFM2.5 Runs Locally on Mac with Strong Performance

LiquidAI released the LFM2.5-2.6B model designed for local agentic workflows, now supported by the open-source macOS app Nativ. On Apple Silicon, it runs a 128K context using only 8.5GB of RAM, achieving highly impressive inference speeds.

2026-08-05 ~ 2026-08-06 · 3 related posts

• LiquidAI's LFM2.5-2.6B Hits 82 tok/s Decode on Mac with 128K Context — helloiamleonie · 2026-08-05
• Nativ local inference hits 11k tok/s prefill with LFM2.5 on M5 Max — helloiamleonie · 2026-08-05
• Nativ brings LiquidAI LFM2.5 to Mac locally: 82 tok/s decode, <8.5GB RAM for 128K context — JosephJacks_ · 2026-08-06