o3
o3 (OpenAI) was released on Apr 16, 2025. In the Epoch AI capability ranking (ECI), it ranks #73 out of 267 models, with 146.87. Against models of the same generation, its best result is on AIME 2024 and 2025 (OTIS): 84.4%, #1 of 52.
Where o3 stands out and where it falls behind, by the numbers
o3 has published results on 33 tests compiled by Epoch AI. The yardstick here is its generation: its position among the models released between Jul 20, 2024 and Jul 15, 2025 that took the same test. Against the whole archive, which includes 2023 models, a recent release lands near the top of almost everything, and that says nothing.
Where it stands out
Tests where it sits in the top third of its own generation.
Where it falls behind
Tests where it sits in the bottom half of its own generation.
See all 33 tests, one by one
| Test | Score | In its generation | Overall | Test record |
|---|---|---|---|---|
| GPQA Diamond | 81.8% | #3 / 54 | #81 / 173 | 96.4% Gemini 4 Argon |
| AIME 2024 and 2025 (OTIS) | 84.4% | #1 / 52 | #74 / 168 | 100.0% GPT-5.5 |
| Dtbench | 84.8% | #2 / 39 | #56 / 148 | 98.9% Claude Opus 5.5 |
| Lmca | 39.70 | #1 / 28 | #47 / 121 | 68.23 Claude Opus 5.5 |
| CritPt | 1.4% | #2 / 22 | #73 / 112 | 32.3% GPT-5.6 Sol |
| WeirdML | 52.4% | #4 / 31 | #43 / 112 | 93.6% GPT-6 Astra |
| Frontiermath Tiers 1 3 V2 | 33.3% | #2 / 7 | #64 / 86 | 93.7% GPT-6 Astra |
| ARC-AGI 2 | 6.5% | #3 / 22 | #52 / 83 | 95.0% GPT-6 Astra |
| ALE-Bench | 934 | #1 / 8 | #44 / 82 | 2,951 GPT-6 Astra |
| ARC-AGI | 60.8% | #2 / 21 | #51 / 81 | 98.5% Claude Fable 5 |
| SimpleQA Verified | 49.4% | #1 / 7 | #28 / 81 | 76.0% Gemini 4 Argon |
| Chess puzzles | 37.0% | #1 / 7 | #36 / 80 | 76.0% GPT-6 Astra |
| Game puzzles | 29.0% | #1 / 4 | #30 / 77 | 84.0% GPT-6 Astra |
| SimpleBench | 53.1% | #4 / 24 | #36 / 74 | 81.9% Claude Fable 5 |
| MATH level 5 | 97.8% | #2 / 42 | #4 / 73 | 98.1% GPT-5 |
| FrontierMath | 18.7% | #3 / 26 | #31 / 65 | 51.7% GPT-5.5 |
| ForecastBench | 62.50 | #1 / 19 | #1 / 62 | 62.50 o3 |
| FrontierMath tier 4 | 2.1% | #5 / 10 (tied with 1) | #34 / 48 | 39.6% GPT-5.5 Pro |
| Hle | 20.3% | #2 / 15 | #20 / 39 | 54.8% GPT-6 Astra |
| Aider Polyglot | 81.30 | #2 / 28 | #3 / 38 | 88.00 GPT-5 |
| Creative writing (Lech Mazur) | 8.39 | #2 / 28 | #5 / 38 | 8.60 GPT-5 |
| EnigmaEval | 13.1% | #1 / 14 | #10 / 36 | 39.3% Claude Fable 5 |
| SWE-bench Verified | 62.3% | #2 / 5 | #27 / 32 | 83.5% Claude Opus 4.7 |
| Fiction.liveBench | 100.0% | #1 / 22 | #1 / 31 | 100.0% o3 |
| METR: time horizon | 120 | #1 / 12 | #10 / 29 | 719 Claude Opus 4.6 |
| Gso | 8.8% | #1 / 8 | #18 / 28 | 88.2% Claude Fable 5.1 |
| DeepResearch Bench | 45.2% | #4 / 7 | #16 / 24 | 55.3% Claude Opus 4.6 |
| GeoBench | 3,930 | #3 / 17 | #6 / 24 | 4,333 Gemini 3 Flash |
| VPCT | 52.0% | #2 / 9 | #7 / 22 | 91.0% Gemini 3 Pro |
| Cl Bench | 17.8% | #1 / 1 | #12 / 19 | 27.9% GPT-5.4 |
| CAD-Eval | 74.0% | #1 / 12 | #1 / 14 | 74.0% o3 |
| GDPval | 30.8% | #1 / 4 | #7 / 11 | 49.7% GPT-5.2 |
| Os World | 23.00 | #3 / 4 | #7 / 8 | 72.10 Claude Sonnet 4.6 |
For each test, the model's best effort level counts, as in the ranking. Source of the scores: Epoch AI, under CC BY 4.0.
What to compare o3 with
The models from other labs closest in capability and released in the same period, at most two per lab. Click any of them to see its page.
| Model | Lab | ECI | ELO | LiveBench | Release |
|---|---|---|---|---|---|
| o3 | OpenAI | 146.87 | n/a | n/a | Apr 16, 2025 |
| Claude Sonnet 4.5 | Anthropic | 146.83 | 1,488 | n/a | Sep 29, 2025 |
| Grok 4 | xAI | 146.43 | n/a | n/a | Jul 9, 2025 |
| DeepSeek-V3.2 | DeepSeek | 146.11 | 1,455 | n/a | Dec 1, 2025 |
| Kimi K2 Thinking | Moonshot | 145.96 | n/a | n/a | Nov 6, 2025 |
| Kimi K2 Thinking Turbo | Moonshot | 145.79 | 1,456 | n/a | Nov 6, 2025 |
| Gemini 2.5 Pro (Jun 2025) | Google DeepMind | 145.31 | 1,507 | n/a | Jun 5, 2025 |
o3 or Claude Sonnet 4.5?
Claude Sonnet 4.5 is the model from another lab closest to o3 in capability. Across the 27 tests both took, this is the score:
| Test | o3 | Claude Sonnet 4.5 |
|---|---|---|
| Chess puzzles | 37.0% | 10.0% |
| Os World | 23.00 | 62.90 |
| EnigmaEval | 13.1% | 6.0% |
| ARC-AGI 2 | 6.5% | 13.6% |
| FrontierMath tier 4 | 2.1% | 4.2% |
| Game puzzles | 29.0% | 17.0% |
| Gso | 8.8% | 14.7% |
| SimpleQA Verified | 49.4% | 30.7% |
| Hle | 20.3% | 13.7% |
| Frontiermath Tiers 1 3 V2 | 33.3% | 23.9% |
Unweighted score: each test counts one point, whether math or coding. The dot marks who did better on each row.
What o3 costs to deliver each task
Price per token is what is on the price sheet; cost per task is what lands on the invoice, because it depends on how many tokens the model spends. Each point is an effort level, published by whoever ran the test and compiled by Epoch AI. The further up and to the left, the more score per dollar.
ARC-AGI 2
The score scale does not start at zero.
ARC-AGI
The score scale does not start at zero.
Other models from OpenAI
The models from the same lab closest in time, newest first.
| Model | Release | ECI | ELO | LiveBench | Access |
|---|---|---|---|---|---|
| gpt-oss-120b | Aug 5, 2025 | 140.00 | 1,385 | n/a | Open weights (unrestricted) |
| o3-pro | Jun 10, 2025 | 147.37 | n/a | n/a | API access |
| o3 | Apr 16, 2025 | 146.87 | n/a | n/a | API access |
| o4-mini | Apr 16, 2025 | 145.60 | n/a | n/a | API access |
| GPT-4.1 | Apr 14, 2025 | 136.77 | n/a | n/a | API access |
| GPT-4.1 mini | Apr 14, 2025 | 135.00 | n/a | n/a | API access |
| GPT-4.1 nano | Apr 14, 2025 | 129.63 | n/a | n/a | API access |
| GPT-4.5 | Feb 27, 2025 | 136.73 | n/a | n/a | API access |
| o3-mini | Jan 31, 2025 | 140.33 | 1,379 | n/a | API access |
Have you used o3? Rate it
Users' opinion, not measurement: the rating is kept apart from the ranking and does not move anyone's position. One vote per browser, no sign-up, and you can change your rating later.
What people ask about o3
Where does o3 rank among AI models?
In the ROO3 AI Benchmark, ordered by the Epoch AI ECI, o3 ranks #73 out of 267 models, with an ECI of 146.87. The ECI is a composite index that combines several hard benchmarks into a single scale. The position changes when a new model comes out, and the page is refreshed six times a day.
Is o3 better than Claude Sonnet 4.5?
Across the 27 tests both models took and Epoch AI compiled, o3 comes out ahead on 16, behind on 11 and tied on 0. It is a test-by-test count with no weighting: a win in math counts the same as a win in coding. To choose between the two, look at the tests that resemble your own use.
What is o3 best at?
Against models of the same generation (released up to 9 months before and up to 3 months after it), the best relative results of o3 are: AIME 2024 and 2025 (OTIS) (#1 of 52); Lmca (#1 of 28); Fiction.liveBench (#1 of 22). The yardstick is its position among the models of that period that took the same test.
Where do the numbers on this page come from?
From the same three sources as the ranking, all openly licensed: Epoch AI (ECI and individual tests, CC BY 4.0), Arena (human-preference ELO, CC BY 4.0) and LiveBench (overall average, CC BY-SA 4.0). ROO3 does not measure any model. When the page shows a figure published by the maker itself, it appears in the written analysis, labeled as the maker's, and never enters the ranking.
Where each number comes from
Epoch AI, "AI Benchmarking Hub". Published at epoch.ai. Data under the CC BY 4.0 licence, reordered by roo3.co.
The Arena (LMArena) leaderboard, from the public lmarena-ai/leaderboard-dataset on Hugging Face, under the CC BY 4.0 licence. We are not affiliated with the Arena.
LiveBench, by the LiveBench team, published at livebench.ai under the CC BY-SA 4.0 license. The overall average is recomputed by roo3.co with the same formula as the website (average of the categories), and those numbers stay under CC BY-SA 4.0. We are not affiliated with LiveBench.
← See the full ranking