Up to four models
Gemini 3.5 Flash vs Gemini 3.1 Pro
On the index (GPT-4 = 100), Gemini 3.5 Flash scores 162 and Gemini 3.1 Pro 161. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
162146–177
WeightsClosed
ReleasedMay 2026
Price in / out$1.5 / $9.0
Context1.05M
Core tests taken7 of 9
161146–177
WeightsClosed
ReleasedFeb 2026
Price in / out$2.0 / $12
Context1.05M
Core tests taken9 of 9
Gemini 3.5 Flash and Gemini 3.1 Pro are too close to call (0 points apart, within the uncertainty), ahead on 4 of 7 shared core tests, and costs 1.3× less.
Test by testthird-party results
Every result as a table
| Test | Gemini 3.5 Flash | Gemini 3.1 Pro |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 92.8% | 94.4% |
| GPQA Diamond (Vals AI) | 92.7% | 95.5% |
| MMLU Pro | 89.5% | 91.0% |
| MMMU Pro | 88.3% | 88.2% |
| EBR-bench | 4.8% | 14.3% |
| GeoBench | 85.0% | 84.0% |
| SimpleBench | 76.7% | 79.6% |
| SimpleQA Verified | 66.2% | 73.5% |
| Maths | ||
| FrontierMath (tiers 1-3) | 62.8% | 59.7% |
| AIME-style maths (OTIS mock) | 95.6% | 95.6% |
| ProofBench v1.1 | 31.0% | 26.0% |
| Chess puzzles | 45.0% | 51.0% |
| FrontierMath (2025 set) | 39.0% | 36.9% |
| FrontierMath tier 4 (2025 set) | 14.6% | 16.7% |
| FrontierMath (tier 4) | 26.8% | 26.8% |
| Coding | ||
| SWE-bench Verified | 79.3% | 75.6% |
| Terminal-Bench 4.0 | 6.1% | 2.5% |
| Code Migration | 26.8% | 17.3% |
| LiveCodeBench | 87.6% | 88.5% |
| ProgramBench | 0.0% | 0.0% |
| SWE-bench | 78.8% | 78.8% |
| Vibe Code Bench v1.1 | 48.7% | 32.0% |
| DeepSWE | 37.4% | 11.7% |
| WeirdML | 62.6% | 72.1% |
| Agents | ||
| APEX-Agents | 27.5% | 35.1% |
| Terminal-Bench-Science | 5.7% | 1.4% |
| DTBench | 94.7% | 97.1% |
| LMCA | 47.1% | 53.8% |
| Mystery game puzzles | 32.0% | 34.0% |
| Professional work | ||
| Finance Agent (v2) | 57.9% | 43.0% |
| CorpFin v2 | 64.7% | 64.5% |
| EMB | 63.5% | 52.6% |
| LegalBench | 83.6% | 87.4% |
| Legal Research Bench | 30.8% | 20.7% |
| MedCode | 55.8% | 59.1% |
| MedScribe | 76.6% | 76.1% |
| MortgageTax | 68.1% | 69.4% |
| Public Benefits Bench v1.1 | 59.5% | 53.8% |
| Tax Agent Bench | 21.6% | 9.3% |
| TaxEval v2 | 74.4% | 72.9% |
| Novel problems | ||
| ARC-AGI-2 | 72.1% | 77.1% |
| ARC-AGI-1 | 92.5% | 98.0% |
Task length · METR
46 sec
Tests at least two of these models have taken. Bold rows are the core tests.