- Humanity's Last Exam (no tools) 18.8
- GPQA diamond, single attempt (pass@1) 84
- AIME 2025, single attempt (pass@1) 86.7
- AIME 2024, single attempt (pass@1) 92
- LiveCodeBench v5, single attempt (pass@1) 70.4
- Aider Polyglot, whole (diff = 68.6%) 74
- SWE-bench verified 63.8
- SimpleQA 52.9
- MMMU, single attempt (pass@1) 81.7
- Vibe-Eval (Reka) 69.4
- Best of MRCR, 128k (average) / MRCR, 1M (pointwise) 94.5
- Global MMLU (Lite) 89.8
Gemini 2.5 Pro Mar 2025
Gemini 2.5 Pro ranks first on 11 of the 65 benchmarks with five or more models scored.
Google DeepMind Released March 25th, 2025 Family Gemini 2.5 The lab's model card Compare with others 70
Benchmarks with a score
11
Ranked first, out of five or more models
26
In the top three
Google DeepMind Mar 2025 Gemini 2.5 Pro
Mar 2025 snapshot Closed weights Flagship
143.9
#103 of 256
Epoch Capabilities Index
Self-Reported Index 143.1 #151 of 404
- Humanity's Last Exam 18.2% #22 /47
- GPQA Diamond 83.8% #84 /229
- LiveCodeBench v5 70.4% #12 /47
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 | |
|---|---|---|---|---|
| MMBench-CN v1.1 | 88.1% | #1 of 178 | Official leaderboard | |
| Vectara Hallucination Leaderboard (HHEM) | 1.10% | #4 of 148 | Official leaderboard | |
| LongBench v2 | 63.3% | #1 of 33 | Official leaderboard | |
| BuzzBench | 71.1 | #1 of 20 | Official leaderboard | |
| LLM Confabulation Leaderboard | 10.8% | #3 of 58 | Official leaderboard | |
| Global MMLU Lite | 89.8% | #1 of 16 | Lab model card | |
| FaithJudge | 7.63% | #3 of 37 | Official leaderboard | |
| MRCR (Michelangelo, original) | 94.5% | #1 of 9 | Lab model card | |
| Arena-Hard v2.0 | 79% | #3 of 23 | Official leaderboard | |
| SimpleQA | 52.9% | #12 of 60 | Lab model card | |
| Fiction.LiveBench | 66.7% | #23 of 57 | Fiction.live leaderboard, via Epoch AI | |
| Humanity's Last Exam | 18.2% | #22 of 47 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| PolyMath | 52.2% | #2 of 38 | Official leaderboard | |
| AIME 2024 | 92% | #9 of 88 | Lab model card | |
| MATH Level 5 | 95.6% | #11 of 94 | Epoch AI run | |
| Putnam-like | 94.5% | #1 of 6 | Official leaderboard | |
| PHYBench | 36.9% | #5 of 28 | Official leaderboard | |
| USAMO 2025 | 24.4% | #2 of 10 | Official leaderboard | |
| Misguided Attention | 55.2% | #12 of 45 | Official leaderboard | |
| GPQA Diamond | 83.8% | #84 of 229 | Epoch AI run | |
| HMMT 2025 | 82.5% | #22 of 53 | Official leaderboard | |
| PutnamBench | 0.45% | #5 of 12 | Official leaderboard | |
| GSM8K Platinum Initial 268 | 99.3% | #8 of 19 | Official leaderboard | |
| UGMathBench | 80.6% | #3 of 7 | Official leaderboard | |
| AIME 2025 | 86.7% | #65 of 143 | Lab model card | |
| SMT 2025 | 84.9% | #20 of 40 | Official leaderboard | |
| BRUMO 2025 | 90% | #22 of 42 | Official leaderboard | |
| SimpleBench | 51.6% | #58 of 106 | Official leaderboard | |
| CMIMC 2025 | 58.1% | #32 of 33 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BigCodeBench-Hard | 33.1% | #6 of 96 | Official leaderboard | |
| Multi-SWE-bench | 14.2% | #1 of 11 | Official leaderboard | |
| LiveCodeBench v5 | 70.4% | #12 of 47 | Lab model card | |
| Aider Polyglot | 74% | #8 of 30 | Lab model card | |
| SWE-bench Verified | 63.8% | #102 of 155 | Lab model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ToolComp | 68.8% | #2 of 32 | Official leaderboard | |
| Galileo Agent Leaderboard | 0.94 | #2 of 29 | Official leaderboard | |
| Step Race | 4 | #9 of 59 | Official leaderboard | |
| VideoGameBench | 0.48% | #1 of 5 | Official leaderboard | |
| BALROG | 43.3% | #11 of 33 | Official leaderboard | |
| VideoGameBench Lite | 1.60% | #3 of 5 | Official leaderboard | |
| LLM Chess | -144 | #88 of 123 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| CadEval | 64% | #2 of 13 | Official leaderboard | |
| HealthBench | 52% | #3 of 8 | Official leaderboard | |
| HealthBench Hard | 18.5% | #3 of 8 | Official leaderboard | |
| Vending-Bench | $789 | #9 of 9 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MMBench-EN v1.1 | 88.5% | #1 of 178 | Official leaderboard | |
| MMStar | 73.6% | #2 of 173 | Official leaderboard | |
| VISTA | 54.6% | #1 of 54 | Official leaderboard | |
| MMMU | 81.7% | #13 of 103 | Lab model card | |
| GeoBench | 84% | #7 of 48 | Official leaderboard | |
| MATH-Vision (MathVision) | 73.3% | #22 of 105 | Official leaderboard | |
| Vibe-Eval | 69.4% | #2 of 9 | Lab model card | |
| VPCT | 48% | #8 of 27 | VPCT leaderboard, via Epoch AI |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| InterCode-CTF | 98.7% | #1 of 4 | Lab model card | |
| Virology Capabilities Test (VCT) | 37.6% | #2 of 7 | Official leaderboard | |
| Hack the Box challenge set (Google DeepMind) | 7.70% | #2 of 4 | Lab model card | |
| In-house CTF suite (Phuong et al. 2024) | 76.9% | #2 of 4 | Lab model card | |
| WMDP Chemistry | 75.3% | #4 of 7 | Lab model card | |
| LAB-Bench ProtocolQA | 69% | #12 of 16 | Lab model card | |
| LAB-Bench CloningScenarios | 44.1% | #9 of 12 | Lab model card | |
| WMDP Biology | 84.3% | #6 of 8 | Lab model card | |
| Cyber key skills benchmark v1 | 33.3% | #3 of 4 | Lab model card | |
| SecureBio VMQA, single choice | 50.8% | #3 of 4 | Lab model card | |
| FORTRESS | 54.9% | #63 of 73 | Official leaderboard | |
| GDM Situational Awareness | 18.2% | #7 of 8 | Lab model card | |
| LAB-Bench SeqQA | 63.9% | #11 of 12 | Lab model card | |
| GDM Stealth | 20% | #6 of 6 | Lab model card |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| OpenVLM Leaderboard | 80.1 | #1 of 173 | Official leaderboard | |
| CAIS Risk Index | 59.0 | #49 of 67 | Official leaderboard | |
| CAIS Text Capabilities Index | 18.8 | #53 of 68 | Official leaderboard |
Its lab's numbers put it at 143.1, in line with 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 140.9 to 145.4. How we compute it.
- MRCR (Michelangelo, original) 94.5
- AIME 2024 92
- Global MMLU Lite 89.8
- AIME 2025 86.7
- GPQA Diamond 84
- MMMU 81.7
- Aider Polyglot 74
- LiveCodeBench v5 70.4
- Vibe-Eval 69.4
- SWE-bench Verified 63.8
- SimpleQA 52.9
- Humanity's Last Exam 18.8
Its lab's numbers next to 4 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.
| Benchmark | Lab says | Board says | Gap |
|---|---|---|---|
| AIME 2025 | 86.7% AIME 2025, single attempt (pass@1) | 88.3% Leaderboard | −1.6 pts |
| Aider Polyglot | 74% Aider Polyglot, whole (diff = 68.6%) | 72.9% Leaderboard, aider, diff-fenced edits | +1.1 pts |
| Humanity's Last Exam | 18.8% Humanity's Last Exam (no tools) | 18.2% Leaderboard | +0.6 pts |
| GPQA Diamond | 84% GPQA diamond, single attempt (pass@1) | 83.8% Epoch AI run | +0.2 pts |
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.
Its system card
Every result in the longer documents the lab published with it. 19 of them aren't in the launch post.
Capabilities
- Humanity's Last Exam 18.8
- GPQA Diamond 84
- AIME 2025 86.7
- AIME 2024 92
- LiveCodeBench v5 70.4
- Aider Polyglot 74
- SWE-bench Verified 63.8
- SimpleQA 52.9
- MMMU 81.7
- Vibe-Eval 69.4
- MRCR (Michelangelo, original) 94.5
- Global MMLU Lite 89.8
Safety
Alignment
- GDM Stealth 20
- GDM Situational Awareness 18.2