K2 Think and Olmo 3 32B Think top the Openness Index at 16 of 100; no other model scores on pre-training data

What the chart shows

Artificial Analysis’s Openness Index scores models from 0 to 100 on how freely their weights can be accessed and licensed and how much developers disclose about training methods and data. Of the 14 models the AI Index shows, K2 Think and Olmo 3 32B Think score highest, at 16 each, and are the only two with any points for pre-training data (2 each).

Most points come from model availability and methodology. Nvidia’s Nemotron 3 Super and Nano score 15, and four models score 7, including gpt-oss-120B and Mistral Large 3. Claude 4.5 Haiku scores 2, all for availability.

Key numbers

How to read it

• Model availability covers access to weights and their licence; the three transparency components cover methodology, pre-training data and post-training data. Components without a bar score zero.
• The 14 models are those the AI Index shows. The full index runs from 0 to 100, but no model shown scores above 16.