OLMo 2 7B: benchmark results
Allen AI's OLMo 2 7B Instruct, a post-trained chat variant of OLMo 2 7B fine-tuned on a Tulu 3-derived dataset through SFT, DPO and RLVR training stages. Provider: Allen AI. Released 2024-11-26. Access: Open.
Aggregate Rating 1355 ± 24, rank #1371 of 1656 rated models, from 30 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| GraphARC - Output Graph Is Tree | 58 | Accuracy (%) of answers to questions about whether the graph | 46.4 |
| AA Humanity's Last Exam | 5.38 | Accuracy (%) | 31.8 |
| GraphARC - Output Graph Component Count | 28 | Accuracy (%) of answers to questions about the number of con | 28.6 |
| MIST (Selective Trust) | 68.1 | Overall Accuracy (%) | 22.7 |
| GraphARC - Output Graph Min Degree | 34 | Accuracy (%) of answers to questions about the minimum node | 21.4 |
| GraphARC - Output Graph Node Count | 19 | Accuracy (%) of answers to questions about the number of nod | 17.9 |
| GraphARC - Output Graph Is Connected | 56 | Accuracy (%) of answers to questions about whether the graph | 14.3 |
| GraphARC - Output Graph Max Degree | 20 | Accuracy (%) of answers to questions about the maximum node | 14.3 |
| Artificial Analysis Intelligence Index | 5.58 | Intelligence Index | 10.3 |
| GraphARC - Output Graph Has Cycles | 44 | Accuracy (%) of answers to questions about whether the graph | 7.1 |
| AA IFBench | 24.42 | Accuracy (%) | 6.7 |
| AA Terminal-Bench Hard | 0 | Accuracy (%) | 5.5 |
Interactive version: theaggregate.ai/model?slug=olmo-2-7b · How It Works · Data refreshed daily, snapshot 2026-10-20.