DeepSeek-R1 (May 2025)
DeepSeek-R1 (May 2025) (DeepSeek) was released on May 28, 2025. In the Epoch AI capability ranking (ECI), it ranks #118 out of 267 models, with 141.27. On Arena, which measures human preference in blind comparisons, it has 1,400 ELO points (#180). Against models of the same generation, its best result is on MATH level 5: 96.6%, #5 of 43.
Where DeepSeek-R1 (May 2025) stands out and where it falls behind, by the numbers
DeepSeek-R1 (May 2025) has published results on 19 tests compiled by Epoch AI. The yardstick here is its generation: its position among the models released between Aug 31, 2024 and Aug 26, 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 19 tests, one by one
| Test | Score | In its generation | Overall | Test record |
|---|---|---|---|---|
| GPQA Diamond | 76.3% | #12 / 60 | #92 / 173 | 96.4% Gemini 4 Argon |
| AIME 2024 and 2025 (OTIS) | 66.4% | #17 / 58 (tied with 1) | #103 / 168 | 100.0% GPT-5.5 |
| CritPt | 1.1% | #6 / 26 | #77 / 112 | 32.3% GPT-5.6 Sol |
| WeirdML | 41.6% | #15 / 38 | #69 / 112 | 93.6% GPT-6 Astra |
| SciCode | 38.3% | #8 / 23 | #83 / 104 | 66.9% Claude Opus 5.5 |
| ARC-AGI 2 | 1.3% | #12 / 26 (tied with 1) | #67 / 83 | 95.0% GPT-6 Astra |
| ALE-Bench | 804 | #4 / 14 | #53 / 82 | 2,951 GPT-6 Astra |
| ARC-AGI | 21.2% | #13 / 25 | #68 / 81 | 98.5% Claude Fable 5 |
| SimpleBench | 40.8% | #11 / 27 | #50 / 74 | 81.9% Claude Fable 5 |
| MATH level 5 | 96.6% | #5 / 43 | #7 / 73 | 98.1% GPT-5 |
| ForecastBench | 60.00 | #13 / 24 | #28 / 62 | 62.50 o3 |
| Aider Polyglot | 71.40 | #8 / 32 | #10 / 38 | 88.00 GPT-5 |
| Creative writing (Lech Mazur) | 8.30 | #8 / 34 (tied with 1) | #9 / 38 | 8.60 GPT-5 |
| BALROG | 34.9% | #3 / 16 | #12 / 32 | 68.3% GPT-6 Astra |
| LiveBench | 71.57 | #6 / 23 | #7 / 32 | 82.35 Gemini 2.5 Pro (Jun 2025) |
| Fiction.liveBench | 33.3% | #24 / 28 | #27 / 31 | 100.0% o3 |
| METR: time horizon | 31.17 | #12 / 15 | #21 / 29 | 719 Claude Opus 4.6 |
| DeepResearch Bench | 35.1% | #9 / 9 | #23 / 24 | 55.3% Claude Opus 4.6 |
| Algotune | 1.70 | #2 / 8 | #7 / 15 | 2.05 GPT-5.2 |
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 DeepSeek-R1 (May 2025) 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 |
|---|---|---|---|---|---|
| o1 | OpenAI | 141.84 | 1,383 | n/a | Dec 17, 2024 |
| Claude Sonnet 4 | Anthropic | 141.67 | 1,414 | n/a | May 22, 2025 |
| DeepSeek-R1 (May 2025) | DeepSeek | 141.27 | 1,400 | n/a | May 28, 2025 |
| Claude 3.7 Sonnet | Anthropic | 141.14 | 1,360 | n/a | Feb 24, 2025 |
| GLM-4.6 | Z.ai (Zhipu AI) | 140.78 | 1,500 | n/a | Sep 30, 2025 |
| o3-mini | OpenAI | 140.33 | 1,379 | n/a | Jan 31, 2025 |
| Grok-3 mini | xAI | 140.33 | 1,394 | n/a | Jun 24, 2025 |
DeepSeek-R1 (May 2025) or Claude 3.7 Sonnet?
Claude 3.7 Sonnet is the model from another lab closest to DeepSeek-R1 (May 2025) in capability. Across the 13 tests both took, this is the score:
| Test | DeepSeek-R1 (May 2025) | Claude 3.7 Sonnet |
|---|---|---|
| METR: time horizon | 31.17 | 60.39 |
| Fiction.liveBench | 33.3% | 53.1% |
| ARC-AGI 2 | 1.3% | 0.9% |
| ARC-AGI | 21.2% | 28.6% |
| DeepResearch Bench | 35.1% | 43.6% |
| Aider Polyglot | 71.40 | 60.40 |
| AIME 2024 and 2025 (OTIS) | 66.4% | 57.8% |
| SimpleBench | 40.8% | 46.4% |
| LiveBench | 71.57 | 76.10 |
| MATH level 5 | 96.6% | 91.2% |
Unweighted score: each test counts one point, whether math or coding. The dot marks who did better on each row.
Other models from DeepSeek
The models from the same lab closest in time, newest first.
| Model | Release | ECI | ELO | LiveBench | Access |
|---|---|---|---|---|---|
| DeepSeek-V4-Pro | Apr 24, 2026 | 149.19 | 1,486 | 71.6 | Open weights (unrestricted) |
| DeepSeek v4 (unknown) | Apr 24, 2026 | 146.26 | n/a | n/a | Open weights (unrestricted) |
| DeepSeek-V3.2 | Dec 1, 2025 | 146.11 | 1,455 | n/a | Open weights (unrestricted) |
| DeepSeek-V3.2-Exp | Sep 29, 2025 | 144.97 | 1,461 | n/a | Open weights (unrestricted) |
| DeepSeek-V3.1 | Aug 21, 2025 | 139.91 | 1,469 | n/a | Open weights (unrestricted) |
| DeepSeek-R1 (May 2025) | May 28, 2025 | 141.27 | 1,400 | n/a | Open weights (unrestricted) |
| DeepSeek-V3 (Mar 2025) | Mar 24, 2025 | 135.93 | 1,346 | n/a | Open weights (restricted use) |
| DeepSeek-R1-Distill-Qwen-32B | Jan 20, 2025 | 137.42 | n/a | n/a | Open weights (unrestricted) |
| DeepSeek-R1-Distill-Qwen-14B | Jan 20, 2025 | 135.44 | n/a | n/a | Open weights (unrestricted) |
Have you used DeepSeek-R1 (May 2025)? 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 DeepSeek-R1 (May 2025)
Where does DeepSeek-R1 (May 2025) rank among AI models?
In the ROO3 AI Benchmark, ordered by the Epoch AI ECI, DeepSeek-R1 (May 2025) ranks #118 out of 267 models, with an ECI of 141.27. 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 DeepSeek-R1 (May 2025) better than Claude 3.7 Sonnet?
Across the 13 tests both models took and Epoch AI compiled, DeepSeek-R1 (May 2025) comes out ahead on 5, behind on 8 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 DeepSeek-R1 (May 2025) 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 DeepSeek-R1 (May 2025) are: MATH level 5 (#5 of 43); BALROG (#3 of 16); Algotune (#2 of 8). 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