Liquid AI ships LFM2.5-24B-A2B: 24B params with 2B active, 184 tok/s on-device agents
JosephJacks_ · x · 2026-10-14
- Liquid AI released LFM2.5-24B-A2B, the largest and final model in its LFM2.5 family: a 24B reasoning model with only 2B active parameters, built for fast on-device agents.
- It beats Qwen3.6-35B-A3B (50% more params) on 5 of 8 agentic benchmarks: 48.09 vs 38.19 on LiveClaw, 27.69 vs 8.72 on AutomationBench.
- Runs up to 1.7x faster than comparable MoEs: 184 tokens/s on Apple M5 Max, up to 21K output tokens/s on one NVIDIA B300.
- Changes vs LFM2-24B-A2B: context extended 32K→131K, vocab doubled 65K→128K, pretraining 17T→33T tokens, large-scale RL, now reasoning-only with explicit chain of thought.
- Base and post-trained weights are open-weight on Hugging Face for local running and fine-tuning.
Related event: LiquidAI Releases LFM2.5-24B-A2B, an On-Device MoE Reasoning Model(13 posts)→
More from Infra
- Jensen Huang joins BG2 podcast: does AI revenue justify capex, '27 bottlenecks — elonmusk · 2026-10-16
- Mistral's on-device infra rack revealed: DGX Spark, Mac Studio and more — gregmushen · 2026-10-16
- NVIDIA and Sweden's Wallenberg launch Sferical One, a shared AI supercomputer — nvidia · 2026-10-16
- Next infra frontier: self-evolving systems that adapt and self-verify — _shreya_s · 2026-10-16
- Xiaomi's MiMo RL dashboard reveals $3.47M spent on just 30 training steps — bycloud · 2026-10-16
- Cerebras pushes disaggregated inference: heterogeneous prefill pools feeding wafer-scale decode — ai · 2026-10-16