Liquid AI ships LFM2.5-24B-A2B: 24B total/2B active local agent model hitting 184 tps on M5 Max
JosephJacks_ · x · 2026-10-14
Liquid AI released LFM2.5-24B-A2B, the largest LFM2.5 model, built specifically for local agents: 24B total parameters with only 2B active per token, 131K context, 33T pretraining tokens, open weights on Hugging Face. Quantized to 14.5GB (Q4KM), it runs at 184 tps on an M5 Max and 114 tps on Ryzen AI Max+ 395. The company claims it beats or matches the 50% larger Qwen3.6-35B-A3B on 5 of 8 agentic/tool benchmarks (AutomationBench 27.69 vs 8.72, LiveClaw 48.09 vs 38.19), and a third-party test showed it winning all 8 terminal tasks against Gemma 4 26B locally, 3.1x faster. Post-training was done inside real agent harnesses (OpenClaw, Hermes Agent, Pi, OpenCode). Note: it still trails Qwen on WebArena and SWE-Bench.
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