tinyanalysis

Guides / Fully open

Best fully open models (open data and code)

Updated November 2026

As of November 2026, the best option is K2-V2 Instruct by LLM360 / MBZUAI (K2), with a Small LM Index of 57.9, followed by OLMo 3.1 32B Think (57.0) and OLMo 3 7B Think (43.4).

Models whose weights, training data and training code are all public, ranked by the Small LM Index.

#ModelLabSmall LM IndexParamsMemory (4-bit)API / 1MLicense
1K2-V2 InstructLLM360 / MBZUAI (K2)57.970B~47 GBself-hostApache-2.0
2OLMo 3.1 32B ThinkAi2 (OLMo)57.032.2B~22 GBself-hostApache-2.0
3OLMo 3 7B ThinkAi2 (OLMo)43.47.3B~7.8 GBself-hostApache-2.0
4SmolLM3 3BHugging Face (SmolLM)30.43.1B~3 GBself-hostApache-2.0
5Apertus 1.5 8BSwiss AI Initiative (Apertus)20.18.9B~7.2 GBself-hostApache-2.0
6EuroLLM 22B InstructEuroLLM (UTTER consortium)19.222.6B~16.5 GBself-hostApache-2.0

Memory covers the weights only; long contexts need more. Scores are mostly lab-reported, so gaps of 1–2 points are noise. See the methodology, or filter the full leaderboard yourself.

Questions

What is the best fully open language model?

As of November 2026, the best option is K2-V2 Instruct by LLM360 / MBZUAI (K2), with a Small LM Index of 57.9, followed by OLMo 3.1 32B Think (57.0) and OLMo 3 7B Think (43.4).

How is this list ranked?

Models whose weights, training data and training code are all public, ranked by the Small LM Index.

Which of the top five needs the least memory?

SmolLM3 3B, at about 3 GB with 4-bit quantization.

Other guides