MiniCPM5-2B: 2B-parameter open model runs agents offline on 2GB RAM, tops sub-4B open models
solyarisoftware · x · 2026-09-10
MiniCPM5-2B is a fully open-source 2B-parameter model that runs agents entirely offline on any laptop or phone with just 2GB RAM.
The poster tested it with a multi-step task — browse Hugging Face, find the Models section, evaluate models on multiple criteria, export a top-15 CSV — and it succeeded, working well even under the Hermes agent framework. In the Artificial Analysis Intelligence Index v4.1.1 cited in its release, it scores 23, ranking #1 among open models under 4B parameters.
The author cites the "Densing Law" from Tsinghua/ModelBest researchers: open pretrained model capability density roughly doubles every 3.5 months — what matters now is intelligence density, not model size.
More from Models
- Testing Astra's self-driven creativity with tree-search prompting: better variety, still lackluster — creatoroff · 2026-09-10
- Dev builds fun AI game with DeepSeek 4.1 Flash, praising its 'gamer temperament' — teortaxesTex · 2026-09-10
- Meta's token share on OpenCode jumps from 3.5% to 45.4% in two weeks on free Muse Spark 1.3 — armand_ruiz · 2026-09-10
- NanoChat implementation details: Apex reports peak memory falling to 140.6 GB — eyishazyer · 2026-09-10
- Mathematician: in 6 months AI went from 'slop' to finishing proofs that take me months — RexDouglass · 2026-09-10
- Blogger quips frontier models' next challenge is cracking Temuka's math — yuvalpi · 2026-09-10