Gemini 3.1 Pro
Gemini 3.1 Pro (Google DeepMind) was released on Feb 19, 2026. In the Epoch AI capability ranking (ECI), it ranks #35 out of 267 models, with 154.67. On Arena, which measures human preference in blind comparisons, it has 1,528 ELO points (#25). On LiveBench, the overall average is 77.0, #22 out of 58 scored models. Against models of the same generation, its best result is on Dtbench: 97.1%, #1 of 57.
Where Gemini 3.1 Pro stands out and where it falls behind, by the numbers
Gemini 3.1 Pro has published results on 43 tests compiled by Epoch AI. The yardstick here is its generation: its position among the models released between May 25, 2025 and May 20, 2026 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 43 tests, one by one
| Test | Score | In its generation | Overall | Test record |
|---|---|---|---|---|
| GPQA Diamond | 94.4% | #2 / 61 | #8 / 173 | 96.4% Gemini 4 Argon |
| AIME 2024 and 2025 (OTIS) | 95.6% | #8 / 61 | #36 / 168 | 100.0% GPT-5.5 |
| Dtbench | 97.1% | #1 / 57 | #8 / 148 | 98.9% Claude Opus 5.5 |
| Lmca | 53.82 | #4 / 54 | #15 / 121 | 68.23 Claude Opus 5.5 |
| CritPt | 17.7% | #5 / 45 | #32 / 112 | 32.3% GPT-5.6 Sol |
| WeirdML | 72.1% | #7 / 53 | #17 / 112 | 93.6% GPT-6 Astra |
| SciCode | 58.7% | #1 / 41 | #14 / 104 | 66.9% Claude Opus 5.5 |
| Frontiermath Tiers 1 3 V2 | 59.6% | #11 / 36 | #38 / 86 | 93.7% GPT-6 Astra |
| ARC-AGI 2 | 77.1% | #4 / 37 | #15 / 83 | 95.0% GPT-6 Astra |
| ALE-Bench | 1,161 | #12 / 47 | #32 / 82 | 2,951 GPT-6 Astra |
| WebDev Arena | 1,447 | #16 / 44 | #51 / 82 | 1,820 Claude Opus 5.5 |
| ARC-AGI | 98.0% | #1 / 37 | #6 / 81 | 98.5% Claude Fable 5 |
| SimpleQA Verified | 73.5% | #1 / 35 | #4 / 81 | 76.0% Gemini 4 Argon |
| Chess puzzles | 51.0% | #5 / 33 | #16 / 80 | 76.0% GPT-6 Astra |
| Game puzzles | 34.0% | #3 / 32 | #22 / 77 | 84.0% GPT-6 Astra |
| ProofBench | 26.0% | #8 / 29 (tied with 1) | #44 / 74 | 100.0% Claude Fable 5.1 |
| SimpleBench | 79.6% | #1 / 33 | #3 / 74 | 81.9% Claude Fable 5 |
| Frontiermath Tier 4 V2 | 26.8% | #9 / 27 (tied with 2) | #34 / 67 | 100.0% GPT-6.1 Sol |
| FrontierMath | 36.9% | #12 / 40 | #13 / 65 | 51.7% GPT-5.5 |
| ForecastBench | 59.00 | #26 / 31 | #38 / 62 | 62.50 o3 |
| Vending Bench 2 | 3,774 | #17 / 33 | #39 / 57 | 15,515 GPT-6 Astra |
| Apex Agents | 35.3% | #7 / 11 | #42 / 50 | 82.4% Gemini 4 Argon |
| FrontierMath tier 4 | 16.7% | #8 / 39 | #9 / 48 | 39.6% GPT-5.5 Pro |
| Hle | 46.4% | #1 / 19 | #3 / 39 | 54.8% GPT-6 Astra |
| Gdp Pdf | 17.0% | #4 / 8 | #27 / 37 | 35.0% GPT-6.1 Sol |
| EnigmaEval | 36.8% | #1 / 18 | #3 / 36 | 39.3% Claude Fable 5 |
| Furniture Assembly | 26.7% | #7 / 8 | #27 / 34 | 83.3% Claude Opus 5.5 |
| Terminal-Bench | 80.2% | #3 / 33 (tied with 1) | #3 / 33 | 84.7% GPT-5.5 |
| BALROG | 57.0% | #2 / 10 | #5 / 32 | 68.3% GPT-6 Astra |
| SWE-bench Verified | 75.6% | #11 / 26 | #12 / 32 | 83.5% Claude Opus 4.7 |
| Blueprint Bench 2 | 26.5% | #4 / 11 | #21 / 31 | 54.4% Gemini 4 Argon |
| METR: time horizon | 384 | #2 / 15 | #2 / 29 | 719 Claude Opus 4.6 |
| Gso | 22.6% | #7 / 14 | #13 / 28 | 88.2% Claude Fable 5.1 |
| Ebr Bench | 14.3% | #5 / 13 (tied with 1) | #18 / 27 | 76.2% GPT-6 Astra |
| DeepSWE | 11.7% | #5 / 5 | #26 / 26 | 74.1% GPT-6 Astra |
| DeepResearch Bench | 47.8% | #9 / 19 | #10 / 24 | 55.3% Claude Opus 4.6 |
| Gbaeval | 0.8% | #8 / 11 | #18 / 22 | 79.6% Claude Opus 5 |
| Cl Bench | 20.8% | #5 / 18 | #5 / 19 | 27.9% GPT-5.4 |
| Algotune | 2.02 | #2 / 13 | #2 / 15 | 2.05 GPT-5.2 |
| Cl Bench Life | 16.9% | #5 / 13 | #5 / 13 | 22.2% GPT-5.5 |
| Posttrainbench | 22.0% | #3 / 4 | #10 / 11 | 41.8% Claude Fable 5 |
| MirrorCode | 8.9% | #4 / 4 | #9 / 9 | 77.4% Claude Opus 5.5 |
| Exploitbench | 26.1% | #3 / 8 | #3 / 8 | 47.4% GPT-5.5 |
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 Gemini 3.1 Pro 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 |
|---|---|---|---|---|---|
| DeepSeek V4 Pro 0813 | DeepSeek | 155.29 | n/a | 77.4 | Aug 13, 2026 |
| Claude Opus 4.6 | Anthropic | 155.27 | 1,550 | 74.5 | Feb 5, 2026 |
| Muse Spark 1.2 | Meta AI | 154.90 | 1,509 | 78.0 | Aug 5, 2026 |
| Gemini 3.1 Pro | Google DeepMind | 154.67 | 1,528 | 77.0 | Feb 19, 2026 |
| DeepSeek V4 Flash 0731 | DeepSeek | 154.49 | n/a | 74.2 | Jul 31, 2026 |
| Grok 4.7 | xAI | 154.37 | 1,453 | 77.4 | Sep 21, 2026 |
| Muse Spark 1.1 | Meta AI | 154.00 | 1,516 | 75.3 | Jul 9, 2026 |
Gemini 3.1 Pro or DeepSeek V4 Flash 0731?
DeepSeek V4 Flash 0731 is the model from another lab closest to Gemini 3.1 Pro in capability. Across the 17 tests both took, this is the score:
| Test | Gemini 3.1 Pro | DeepSeek V4 Flash 0731 |
|---|---|---|
| SimpleQA Verified | 73.5% | 33.6% |
| ProofBench | 26.0% | 56.0% |
| Chess puzzles | 51.0% | 32.0% |
| SimpleBench | 79.6% | 61.1% |
| Lmca | 53.82 | 41.73 |
| ARC-AGI 2 | 77.1% | 61.4% |
| SciCode | 58.7% | 49.9% |
| WeirdML | 72.1% | 63.0% |
| ALE-Bench | 1,161 | 1,306 |
| ARC-AGI | 98.0% | 89.0% |
Unweighted score: each test counts one point, whether math or coding. The dot marks who did better on each row.
Other models from Google DeepMind
The models from the same lab closest in time, newest first.
| Model | Release | ECI | ELO | LiveBench | Access |
|---|---|---|---|---|---|
| Gemini 3.6 Flash | Jul 21, 2026 | 154.24 | 1,530 | 73.6 | API access |
| Gemini 3.5 Flash-Lite | Jul 21, 2026 | 145.15 | 1,467 | 63.9 | API access |
| Gemini 3.5 Flash | May 19, 2026 | 154.38 | 1,526 | 74.6 | API access |
| Gemma 4 31B IT | Apr 2, 2026 | 142.86 | n/a | n/a | Open weights (restricted use) |
| Gemma 4 26B A4B | Apr 2, 2026 | 141.81 | 1,495 | n/a | Open weights (unrestricted) |
| Gemini 3.1 Pro | Feb 19, 2026 | 154.67 | 1,528 | 77.0 | API access |
| Gemini 3 Flash | Dec 17, 2025 | 151.80 | 1,511 | n/a | API access |
| Gemini 3 Pro | Nov 18, 2025 | 152.90 | 1,523 | n/a | API access |
| Gemini 2.5 Flash (Sep 2025) | Sep 25, 2025 | 143.00 | 1,450 | n/a | API access |
Have you used Gemini 3.1 Pro? 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 Gemini 3.1 Pro
Where does Gemini 3.1 Pro rank among AI models?
In the ROO3 AI Benchmark, ordered by the Epoch AI ECI, Gemini 3.1 Pro ranks #35 out of 267 models, with an ECI of 154.67. 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 Gemini 3.1 Pro better than DeepSeek V4 Flash 0731?
Across the 17 tests both models took and Epoch AI compiled, Gemini 3.1 Pro comes out ahead on 14, behind on 2 and tied on 1. 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 Gemini 3.1 Pro 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 Gemini 3.1 Pro are: Dtbench (#1 of 57); SciCode (#1 of 41); ARC-AGI (#1 of 37). 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