Model card
Gemini 3 Pro.
GoogleapiUndisclosed params1 current SOTA
Google flagship model.
§ 02 · Benchmarks
Every benchmark Gemini 3 Pro has a recorded score for.
| # | Benchmark | Area · Task | Metric | Value | Rank | Date | Source |
|---|---|---|---|---|---|---|---|
| 01 | GPQA Diamond | Reasoning · Multi-step Reasoning | accuracy | 91.9% | #1 | — | source ↗ |
| 02 | LiveCodeBench Pro | Computer Code · Code Generation | elo | 2439.00 | #2 | — | source ↗ |
| 03 | SWE-bench Verified | Agentic AI · Autonomous Coding | pct_resolved | 78.8% | #2 | — | source ↗ |
| 04 | MMMU-Pro | Multimodal · Visual Question Answering | accuracy | 80.0% | #3 | 2026-01-15 | source ↗ |
| 05 | Tau2-Bench | Agentic AI · Tool Use | pass_rate | 69.0% | #3 | 2025-11-18 | source ↗ |
| 06 | MMLU | Reasoning · Commonsense Reasoning | accuracy | 91.4% | #6 | 2026-01-01 | source ↗ |
| 07 | HLE | Reasoning · Multi-step Reasoning | accuracy | 38.3% | #6 | — | |
| 08 | SWE-bench | Computer Code · Code Generation | resolve-rate-agentic | 77.4% | #8 | 2026-01-01 | source ↗ |
| 09 | OmniDocBench | Computer Vision · Document Parsing | composite | 90.3% | #8 | — | source ↗ |
| 10 | SWE-bench | Computer Code · Code Generation | resolve-rate | 77.4% | #10 | 2026-01-01 | source ↗ |
| 11 | SWE-bench | Computer Code · Code Generation | resolve-rate-agentic | 76.2% | #12 | 2025-12-01 | |
| 12 | SWE-Bench Verified | Computer Code · Code Generation | resolve-rate | 76.2% | #12 | — | source ↗ |
| 13 | SWE-bench Verified | Agentic AI · SWE-bench | resolve-rate | 76.2% | #19 | — | source ↗ |
Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.
§ 03 · Strengths by area
Where Gemini 3 Pro actually performs.
§ 04 · Papers
1 paper with results for Gemini 3 Pro.
- 2023-10-10· Computer Code· 1 result
SWE-bench: Can Language Models Resolve Real-World GitHub Issues?
Carlos E. Jimenez, John Yang, Alexander Wettig, Shunyu Yao et al.
§ 05 · Related models
Other Google models scored on Codesota.
§ 06 · Sources & freshness
Where these numbers come from.
editorial
4
results
google-blog
2
results
livecodebench-pro-official
1
result
artificialanalysis.ai
1
result
codesota-shadow-mmlu
1
result
live-swe-agent
1
result
paddleocr-paper
1
result
swebench-leaderboard
1
result
google-internal
1
result
6 of 13 rows marked verified. · first result 2025-11-18, latest 2026-01-15.