Model comparison
Gemini 3.1 Pro Preview vs Gemini 3 Pro
Gemini 3.1 Pro Preview is the stronger model overall, scoring 56.7 to 54.7 on the Noometry Index.
Last verified . 51 shared benchmarks.
GoogleSummary
- They share 51 benchmarks with published results for both. Gemini 3.1 Pro Preview scores higher in 6 categories and Gemini 3 Pro in 4 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Gemini 3.1 Pro Preview leads 71.7 to 52.5.
- The biggest single-benchmark swing is ARC-AGI-2: 77.1% for Gemini 3.1 Pro Preview and 31.1% for Gemini 3 Pro.
Side by side
| Gemini 3.1 Pro Preview | Gemini 3 Pro | |
|---|---|---|
| Provider | ||
| Noometry Index | 56.7 | 54.7 |
| Released | 2026-02-19 | 2025-11-18 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | — |
| Max output | 66K | — |
| Input $ / M tokens | $2 | — |
| Output $ / M tokens | $12 | — |
| Results tracked | 71 | 67 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Gemini 3 Pro leads
Gemini 3.1 Pro Preview: 42.7 (#88), Gemini 3 Pro: 52.1 (#33)
| Benchmark | Gemini 3.1 Pro Preview | Gemini 3 Pro |
|---|---|---|
| SWE-bench Verified | 75.6% | 72.9% |
| LMArena WebDev | 1446 | 1439 |
| GSO | 22.6% | 18.6% |
| WeirdML | 72.1% | 69.9% |
| LMArena Coding | 1483 | 1482 |
| ALE-Bench | 1,161 | 1,177 |
| AlgoTune | 2.02 | 1.83 |
| DeepSWE | 11.7% | — |
| SWE-bench Verified (bash only) | — | 74.2% |
| SWE-bench Multilingual | — | 68.7% |
| SciCode | 58.9% | — |
| MirrorCode | 8.9% | — |
Agentic & Tool Use Gemini 3 Pro leads
Gemini 3.1 Pro Preview: 37.7 (#33), Gemini 3 Pro: 40.5 (#21)
| Benchmark | Gemini 3.1 Pro Preview | Gemini 3 Pro |
|---|---|---|
| Terminal-Bench | 80.2% | 69.4% |
| τ²-bench Banking | 26% | 18% |
| DeepResearch Bench | 47.8% | 46.3% |
| BALROG | 57% | 58.1% |
| LMArena Search | 1211 | 1207 |
| METR Time Horizons | 77% | 71% |
| Vending-Bench 2 | 3,774 | 5,478 |
| APEX-Agents | 35.3% | — |
| Berkeley Function Calling Leaderboard | — | 72.5% |
| GDPval | — | 40.3% |
| Remote Labor Index | — | 1.3% |
| τ²-bench Airline | — | 80.5% |
| τ²-bench Retail | — | 75.9% |
| τ²-bench Telecom | — | 91% |
| PostTrainBench | 22% | — |
| ExploitBench | 26.1% | — |
| GBAEval | 0.8% | — |
| GDP.pdf | 17% | — |
Reasoning Gemini 3.1 Pro Preview leads
Gemini 3.1 Pro Preview: 71.7 (#9), Gemini 3 Pro: 52.5 (#28)
| Benchmark | Gemini 3.1 Pro Preview | Gemini 3 Pro |
|---|---|---|
| ARC-AGI-2 | 77.1% | 31.1% |
| SimpleBench | 79.6% | 76.4% |
| NYT Connections (extended) | 97.4% | 94.4% |
| ARC-AGI-1 | 98% | 75% |
| CritPt | 17.7% | 6.9% |
| Chess Puzzles | 55% | 31% |
| EnigmaEval | 36.8% | 18.2% |
| LMArena Hard Prompts | 1484 | 1481 |
| Epoch Capabilities Index | 154.92 | 152.96 |
| ForecastBench | 59 | 61.2 |
| Kagi LLM Benchmark | — | 80.1% |
| Thematic Generalization | 79.4% | — |
| EBR-Bench | 14.3% | — |
| Mystery Game Puzzles | 34% | — |
| DTBench | 97.1% | — |
| LMCA | 53.8% | — |
Math Gemini 3.1 Pro Preview leads
Gemini 3.1 Pro Preview: 61.8 (#28), Gemini 3 Pro: 48.8 (#50)
| Benchmark | Gemini 3.1 Pro Preview | Gemini 3 Pro |
|---|---|---|
| MathArena Final-Answer Competitions | 86.5% | 67% |
| OTIS Mock AIME 2024-2025 | 95.6% | 91.4% |
| ProofBench | 26% | 20% |
| LMArena Math | 1485 | 1475 |
| FrontierMath (Feb 2025 set) | 36.9% | 37.6% |
| FrontierMath Tier 4 (v1) | 16.7% | 18.8% |
| FrontierMath (Tiers 1-3) | 59.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| Omni-MATH | — | 55.5% |
Knowledge Gemini 3.1 Pro Preview leads
Gemini 3.1 Pro Preview: 71.8 (#3), Gemini 3 Pro: 64.3 (#15)
| Benchmark | Gemini 3.1 Pro Preview | Gemini 3 Pro |
|---|---|---|
| GPQA Diamond | 94.4% | 92.6% |
| Humanity's Last Exam | 46.4% | 37.5% |
| Vectara Hallucination Rate | 10.4% | 13.6% |
| LMArena Expert | 1484 | 1475 |
| SimpleQA Verified | 73.5% | — |
| MMLU-Pro | — | 90.3% |
| GPQA (HELM) | — | 80.3% |
Multimodal Gemini 3 Pro leads
Gemini 3.1 Pro Preview: 38.1 (#60), Gemini 3 Pro: 56.9 (#1)
| Benchmark | Gemini 3.1 Pro Preview | Gemini 3 Pro |
|---|---|---|
| LMArena Vision | 1296 | 1305 |
| LMArena Document | 1444 | 1434 |
| GeoBench | — | 84% |
| VPCT | — | 91% |
| Blueprint-Bench 2 | 26.5% | — |
| Furniture Assembly | 26.7% | — |
Multilingual Too close to call
Gemini 3.1 Pro Preview: 57.1 (#11), Gemini 3 Pro: 57.1 (#14)
| Benchmark | Gemini 3.1 Pro Preview | Gemini 3 Pro |
|---|---|---|
| LMArena Non-English | 1475 | 1474 |
| LMArena Chinese | 1528 | 1522 |
| LMArena French | 1488 | 1492 |
| LMArena German | 1491 | 1517 |
| LMArena Japanese | 1494 | 1513 |
| LMArena Korean | 1458 | 1449 |
| LMArena Russian | 1496 | 1493 |
| LMArena Spanish | 1480 | 1468 |
Instruction Following Too close to call
Gemini 3.1 Pro Preview: 77.1 (#31), Gemini 3 Pro: 76.4 (#42)
| Benchmark | Gemini 3.1 Pro Preview | Gemini 3 Pro |
|---|---|---|
| LMArena Instruction Following | 1465 | 1458 |
| IFEval | — | 87.7% |
Long Context Gemini 3.1 Pro Preview leads
Gemini 3.1 Pro Preview: 47.3 (#17), Gemini 3 Pro: 44.0 (#75)
| Benchmark | Gemini 3.1 Pro Preview | Gemini 3 Pro |
|---|---|---|
| CL-bench | 20.8% | 15.8% |
| LMArena Longer Query | 1482 | 1472 |
| CL-bench Life | 16.9% | — |
Writing & Preference Too close to call
Gemini 3.1 Pro Preview: 66.3 (#36), Gemini 3 Pro: 66.6 (#34)
| Benchmark | Gemini 3.1 Pro Preview | Gemini 3 Pro |
|---|---|---|
| LMArena Text | 1480 | 1479 |
| LMArena Creative Writing | 1482 | 1482 |
| EQ-Bench Creative Writing | 1491 | 1525 |
| LMArena Multi-Turn | 1486 | 1484 |
| WildBench | — | 85.9% |
| EQ-Bench 4 | 1142 | — |
Frequently asked questions
Is Gemini 3.1 Pro Preview better than Gemini 3 Pro?
Gemini 3.1 Pro Preview is the stronger model overall, scoring 56.7 to 54.7 on the Noometry Index.
Is Gemini 3.1 Pro Preview or Gemini 3 Pro better for coding?
Gemini 3 Pro scores higher on coding benchmarks: 52.1 versus 42.7 in the Noometry coding category.
How many benchmarks do Gemini 3.1 Pro Preview and Gemini 3 Pro share?
51 benchmarks have published results for both models. Gemini 3.1 Pro Preview has 71 scored results on Noometry and Gemini 3 Pro has 67.