Model
Gemini 3 Pro
Knowledge & Reasoning
Knowledge & Reasoning| Benchmark | Score | Metric | Harness | Source | Confidence | Link |
|---|---|---|---|---|---|---|
| Balrog | 58.1 | average_progress | N/A | Independent | medium | balrogai.com |
| CL-bench | 15.8 | overall | N/A | Independent | low | epoch.ai |
| Contradiction Benchmark for Review Evaluation | 99.9 | accuracy | N/A | Independent | medium | arxiv.org |
| CritPt | 6.9 | accuracy | N/A | Independent | low | epoch.ai |
| EnigmaEval | 18.2 | accuracy | N/A | Independent | medium | labs.scale.com |
| GDPval | 40.3 | win_rate | N/A | Independent | low | epoch.ai |
| GPQA Diamond | 92.6 | accuracy | N/A | Independent | high | epoch.ai |
| GPQA Diamond | 87.0 | accuracy | N/A | Self-reported | low | artificialanalysis.ai |
| HLE | 45.8 | accuracy | N/A | Self-reported | low | lastexam.ai |
| HLE | 37.5 | accuracy | N/A | Independent | low | epoch.ai |
| LMArena | 1486.0 | elo | N/A | Self-reported | low | lmarena.ai |
| MMLU-Pro | 85.0 | accuracy | N/A | Self-reported | low | artificialanalysis.ai |
Math
Math| Benchmark | Score | Metric | Harness | Source | Confidence | Link |
|---|---|---|---|---|---|---|
| AIME 2025 | 96.0 | accuracy | N/A | Self-reported | low | kaggle.com |
| MATH-500 | 91.0 | accuracy | N/A | Self-reported | low | kaggle.com |
Coding
Coding| Benchmark | Score | Metric | Harness | Source | Confidence | Link |
|---|---|---|---|---|---|---|
| ARCS | 44.0 | ex_e2e | N/A | Independent | medium | arxiv.org |
| GSO | 18.6 | score_opt@1 | OpenHands | Independent | medium | gso-bench.github.io |
| LiveCodeBench | 71.0 | pass@1 | N/A | Self-reported | low | livecodebench.github.io |
| QTestBench | 1.0 | test_quality | N/A | Independent | medium | arxiv.org |
| SWE-bench Verified | 75.0 | resolved_rate | N/A | Self-reported | low | swebench.com |
| SWE-bench Verified | 72.9 | resolved_rate | N/A | Independent | high | epoch.ai |
Agentic / Tool Use
Agentic / Tool Use| Benchmark | Score | Metric | Harness | Source | Confidence | Link |
|---|---|---|---|---|---|---|
| DeepResearch Bench | 46.3 | average_score | N/A | Independent | medium | evals.futuresearch.ai |
| MisVisAgentBench | 45.7 | pra | N/A | Independent | medium | arxiv.org |
| τ²-bench | 65.8 | pass^1 | N/A | Self-reported | low | github.com |
| Terminal-Bench | 69.4 | accuracy | Ante | Independent | medium | tbench.ai |
| Terminal-Bench | 65.2 | accuracy | SageAgent | Independent | medium | tbench.ai |
| Terminal-Bench | 62.2 | accuracy | CodeBrain-1.5 | Independent | medium | tbench.ai |
| Terminal-Bench | 62.2 | accuracy | CodeBrain-1 | Independent | low | epoch.ai |
| Terminal-Bench | 61.8 | accuracy | II-Agent | Independent | medium | tbench.ai |
| Terminal-Bench | 61.1 | accuracy | Droid | Independent | medium | tbench.ai |
| Terminal-Bench | 56.9 | accuracy | Terminus 2 | Independent | medium | tbench.ai |
| Terminal-Bench | 56.0 | accuracy | Letta Code | Independent | medium | tbench.ai |
Multimodal
Multimodal| Benchmark | Score | Metric | Harness | Source | Confidence | Link |
|---|---|---|---|---|---|---|
| GeoBench | 84.0 | acw_country_% | N/A | Independent | medium | geobench.org |
| SQUARE-Bench | 91.0 | overall | N/A | Independent | medium | arxiv.org |
Instruction Following
Instruction Following| Benchmark | Score | Metric | Harness | Source | Confidence | Link |
|---|---|---|---|---|---|---|
| IFEval | 82.0 | prompt_strict_accuracy | N/A | Self-reported | low | artificialanalysis.ai |
Abstract Reasoning
Abstract Reasoning| Benchmark | Score | Metric | Harness | Source | Confidence | Link |
|---|---|---|---|---|---|---|
| ARC-AGI-2 | 84.6 | accuracy | N/A | Self-reported | low | arcprize.org |
| ARC-AGI-2 | 31.1 | score | N/A | Independent | low | epoch.ai |
Reproduced by the community
Loading community runs…