Ai2
What its model cards measure
Tall ticks are Ai2 model cards that list the benchmark in their headline table. Gray dots are cards that leave it out. Rows are the benchmarks the lab listed most, at least twice each, in the order it started using them.
Model cards
| Date | Model | Document | Benchmarks |
|---|---|---|---|
| 5 May 2026 | MolmoAct 2 | MolmoAct 2 (Ai2 blog) | 3 |
| 24 Mar 2026 | MolmoWeb 8B | MolmoWeb (Ai2 blog) | 4 |
| 27 Jan 2026 | SERA-32B | Open Coding Agents: SERA (Ai2 blog) | 1 |
| 12 Dec 2025 | Olmo 3.1 32B Think | Olmo-3.1-32B-Think model card (Hugging Face) | 17 |
| 20 Nov 2025 | Olmo 3 32B Think | Olmo-3-32B-Think model card (Hugging Face) | 17 |
| 29 Apr 2025 | OLMo 2 1B Instruct | OLMo-2-0425-1B-Instruct model card (Hugging Face) | 10 |
| 13 Mar 2025 | OLMo 2 32B Instruct | OLMo-2-0325-32B-Instruct model card (Hugging Face) | 10 |
| 30 Jan 2025 | Tulu 3 405B | Llama-3.1-Tulu-3-405B model card (Hugging Face) | 10 |
Every Ai2 model
- MolmoAct 2 May 2026
- MolmoWeb 8B Mar 2026
- SERA-32B Jan 2026
- Molmo2-8B Dec 2025
- Olmo 3.1 32B Think Dec 2025
- Olmo 3 32B Think Nov 2025
- OLMo 2 1B Instruct Apr 2025
- OLMo 2 32B Instruct Mar 2025
- Tulu 3 405B Jan 2025
- Tulu 3 (Tülu 3) 70B Nov 2024
- Molmo-7B-D Sep 2024
- Molmo-7B-O Sep 2024
- Molmo-72B Sep 2024
- MolmoE-1B Sep 2024
- OLMo-7B-Instruct Feb 2024
- Tulu 2 DPO 70B Nov 2023