All tasks / Models
Molmo2-8B
Ai2's fully open vision-language models (weights, data and code) for images, video, pointing and tracking.
- Allen Institute for AI (Ai2)
- Open weights
- Released Dec 2025
Where it ranks
- CImage understandingModels · Local · 1,112 on LMArena Vision Arena (overall, style control)
Facts
- Vendor
- Allen Institute for AI (Ai2)
- Family
- Molmo
- Released
- Dec 2025
- Weights
- Open: download and run it yourself
- License
- Apache-2.0
- Parameters
- 8.7B
- GPU memory to run locally
- ~6.5 GB (8.66B (HF safetensors), Qwen3-8B + SigLIP base; 4-bit ≈ 8.7 × 0.57 + 1.5 ≈ 6.5 GB, estimate.)
- Access
- Free (open weights)
- Last verified
- 19 Oct 2026
Price from huggingface.co, 19 Oct 2026.
Benchmark results
| Benchmark | Result | Entry on the leaderboard | As of | Source |
|---|---|---|---|---|
| Artificial Analysis MMMU-Pro (independent evaluation) | 37.5% | Molmo2-8BRaw fraction 0.3746 from the model data embedded in https://artificialanalysis.ai/leaderboards/models, converted to percent; rank 118 of 124 entries with an MMMU-Pro result; no last-updated date shown. | 19 Oct 2026 | artificialanalysis.ai |
| LMArena Vision Arena (overall, style control) | 1,112 | molmo-2-8bFrom LMArena's official leaderboard-dataset (vision_style_control, split latest, category overall, published 2026-10-09), which mirrors https://arena.ai/leaderboard/vision. Rank 133 of 165; 1189 votes; 95% CI 1092.3-1131.8. | 9 Oct 2026 | huggingface.co |