- MMLU-Pro 79.1
- LiveCodeBench (v5) (UI 10/01/2024-02/01/2025) 36
- Bird-SQL (Dev) 59.3
- GPQA (diamond) 64.7
- SimpleQA 44.3
- FACTS Grounding 82.8
- Global MMLU (Lite) 86.5
- MATH 91.8
- HiddenMath 65.2
- MRCR (1M) 74.7
- MMMU 72.7
- CoVoST2 (21 lang), BLEU 40.6
- EgoSchema (test) 71.9
Gemini 2.0 Pro
Gemini 2.0 Pro ranks first on 2 of the 51 benchmarks with five or more models scored.
51
Benchmarks with a score
2
Ranked first, out of five or more models
7
In the top three
Google DeepMind Feb 2025 Gemini 2.0 Pro
Closed weights
135.3
#139 of 256
Epoch Capabilities Index
Self-Reported Index 133.1 #248 of 404
- GPQA Diamond 65.7% #128 /229
- LiveCodeBench v5 36% #39 /47
- Bird-SQL 59.3% #4 /13
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Vectara Hallucination Leaderboard (HHEM) | 0.80% | #3 of 148 | Official leaderboard | |
| Creation-MMBench | 4.48 | #1 of 19 | Official leaderboard | |
| MMBench-CN v1.1 | 83.5% | #23 of 178 | Official leaderboard | |
| SimpleQA | 44.3% | #17 of 60 | Lab model card | |
| MRCR (Michelangelo, original) | 74.7% | #3 of 9 | Lab model card | |
| Global MMLU Lite | 86.5% | #6 of 16 | Lab model card | |
| MMLU-Pro | 79.1% | #84 of 167 | Lab model card | |
| LLM Confabulation Leaderboard | 18.4% | #34 of 58 | Official leaderboard | |
| FACTS Grounding | 82.8% | #10 of 15 | Lab model card | |
| Fiction.LiveBench | 41.7% | #50 of 57 | Fiction.live leaderboard, via Epoch AI |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MATH | 91.8% | #7 of 99 | Lab model card | |
| HiddenMath | 65.2% | #2 of 11 | Lab model card | |
| MATH Level 5 | 83.5% | #24 of 94 | Epoch AI run | |
| GPQA Diamond | 65.7% | #128 of 229 | Epoch AI run | |
| 3SAT-Search | 25.1% | #7 of 11 | Official leaderboard | |
| TSP-Opt | 20.5% | #7 of 11 | Official leaderboard | |
| Misguided Attention | 34.6% | #31 of 45 | Official leaderboard | |
| AIME 2025 | 27.5% | #50 of 53 | Official leaderboard | |
| GSM8K Platinum Initial 268 | 97.4% | #18 of 19 | Official leaderboard | |
| HMMT 2025 | 7.50% | #51 of 53 | Official leaderboard | |
| GSM8K-Platinum | 98.5% | #8 of 8 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BigCodeBench-Hard | 31.8% | #12 of 96 | Official leaderboard | |
| Bird-SQL | 59.3% | #4 of 13 | Lab model card | |
| Aider Polyglot | 35.6% | #28 of 46 | Official leaderboard | |
| Spider 2.0 | 13.9% | #19 of 23 | Official leaderboard | |
| LiveCodeBench v5 | 36% | #39 of 47 | Lab model card | |
| BIRD-CRITIC | 21% | #8 of 9 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ToolComp | 61.5% | #13 of 32 | Official leaderboard | |
| Step Race | 1 | #39 of 59 | Official leaderboard | |
| Elimination Game | 3 | #47 of 56 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| FinanceReasoning | 86.3% | #1 of 6 | Official leaderboard | |
| TaxEval v2 | 68.2% | #92 of 129 | Official leaderboard | |
| CorpFin v2 | 43.4% | #108 of 119 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MM-Vet | 70.4% | #19 of 164 | Official leaderboard | |
| MMStar | 68.5% | #21 of 173 | Official leaderboard | |
| MMBench | 84.3% | #20 of 155 | Official leaderboard | |
| CoVoST2 | 40.6 | #2 of 14 | Lab model card | |
| VHELM | 72.4 | #6 of 41 | Official leaderboard | |
| MMBench-EN v1.1 | 82.4% | #35 of 178 | Official leaderboard | |
| MMMU | 72.7% | #27 of 136 | Official leaderboard | |
| GeoBench | 77.5% | #16 of 48 | Official leaderboard | |
| VisuLogic | 28% | #6 of 17 | Official leaderboard | |
| MATH-Vision (MathVision) | 48.1% | #54 of 105 | Official leaderboard | |
| VISTA | 43.3% | #29 of 54 | Official leaderboard | |
| EgoSchema | 71.9% | #7 of 12 | Lab model card |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Video-SafetyBench | 1.70 | #2 of 15 | Official leaderboard | |
| AIR-Bench 2024 | 68.4% | #37 of 82 | Official leaderboard | |
| Anthropic Red Team | 98.7% | #44 of 75 | Official leaderboard | |
| HELM Safety | 90.5 | #44 of 75 | Official leaderboard | |
| SimpleSafetyTests | 97.5% | #53 of 75 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| OpenVLM Leaderboard | 73.3 | #19 of 173 | Official leaderboard |
Its lab's numbers put it at 133.1, 2.2 points below its ECI
The Self-Reported Index fits one score to the 12 capability results Google DeepMind reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 129.9 to 136.3. How we compute it.
- MATH 91.8
- Global MMLU Lite 86.5
- FACTS Grounding 82.8
- MMLU-Pro 79.1
- MRCR (Michelangelo, original) 74.7
- MMMU 72.7
- EgoSchema 71.9
- HiddenMath 65.2
- GPQA Diamond 64.7
- Bird-SQL 59.3
- SimpleQA 44.3
- LiveCodeBench v5 36
Its lab's numbers next to 3 leaderboards
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
Google DeepMind's models over time
Google DeepMind's models in color, against every other model on the Epoch Capabilities Index, by release date.
Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.