Model comparison
Gemini 2.5 Pro vs Gemini 3 Pro
Gemini 3 Pro is the stronger model overall, scoring 54.8 to 45.0 on the Noometry Index.
Last verified . 58 shared benchmarks.
GoogleSummary
- They share 58 benchmarks with published results for both. Gemini 2.5 Pro scores higher in 1 category and Gemini 3 Pro in 9 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Gemini 3 Pro leads 52.5 to 28.8.
- The biggest single-benchmark swing is VPCT: 48% for Gemini 2.5 Pro and 91% for Gemini 3 Pro.
Side by side
| Gemini 2.5 Pro | Gemini 3 Pro | |
|---|---|---|
| Provider | ||
| Noometry Index | 45.0 | 54.8 |
| Released | 2025-03-25 | 2025-11-18 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | — |
| Max output | 66K | — |
| Input $ / M tokens | $1.25 | — |
| Output $ / M tokens | $10 | — |
| Results tracked | 78 | 67 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Gemini 3 Pro leads
Gemini 2.5 Pro: 42.4 (#101), Gemini 3 Pro: 51.6 (#39)
| Benchmark | Gemini 2.5 Pro | Gemini 3 Pro |
|---|---|---|
| SWE-bench Verified | 57.6% | 72.9% |
| SWE-bench Verified (bash only) | 53.6% | 74.2% |
| LMArena WebDev | 1227 | 1440 |
| GSO | 3.9% | 18.6% |
| WeirdML | 54% | 69.9% |
| LMArena Coding | 1452 | 1481 |
| ALE-Bench | 785.52 | 1,177 |
| AlgoTune | 1.51 | 1.83 |
| Aider Polyglot | 83.1% | — |
| SWE-bench Multilingual | — | 68.7% |
| SciCode | 42.8% | — |
| LiveBench Coding | 85.9% | — |
| CadEval | 64% | — |
Agentic & Tool Use Gemini 3 Pro leads
Gemini 2.5 Pro: 29.2 (#88), Gemini 3 Pro: 40.6 (#23)
| Benchmark | Gemini 2.5 Pro | Gemini 3 Pro |
|---|---|---|
| Terminal-Bench | 32.6% | 69.4% |
| GDPval | 23.3% | 40.3% |
| Remote Labor Index | 0.8% | 1.3% |
| τ²-bench Banking | 13.7% | 18% |
| DeepResearch Bench | 42.8% | 46.3% |
| BALROG | 43.3% | 58.1% |
| LMArena Search | 1142 | 1207 |
| METR Time Horizons | 55.4% | 71% |
| Vending-Bench 2 | 573.64 | 5,478 |
| Berkeley Function Calling Leaderboard | — | 72.5% |
| TheAgentCompany | 30.3% | — |
| τ²-bench Airline | — | 80.5% |
| τ²-bench Retail | — | 75.9% |
| τ²-bench Telecom | — | 91% |
Reasoning Gemini 3 Pro leads
Gemini 2.5 Pro: 28.8 (#99), Gemini 3 Pro: 52.5 (#31)
| Benchmark | Gemini 2.5 Pro | Gemini 3 Pro |
|---|---|---|
| ARC-AGI-2 | 4.9% | 31.1% |
| SimpleBench | 62.4% | 76.4% |
| Kagi LLM Benchmark | 70.3% | 80.1% |
| ARC-AGI-1 | 41% | 75% |
| CritPt | 2% | 6.9% |
| Chess Puzzles | 20% | 31% |
| EnigmaEval | 5.6% | 18.2% |
| LMArena Hard Prompts | 1455 | 1480 |
| Epoch Capabilities Index | 145.32 | 152.92 |
| ForecastBench | 61.3 | 61.2 |
| NYT Connections (extended) | — | 94.4% |
| LiveBench Reasoning | 89.8% | — |
| DTBench | 82.4% | — |
| LiveBench Data Analysis | 79.9% | — |
| LMCA | 34.8% | — |
| LiveBench | 82.3% | — |
Math Gemini 3 Pro leads
Gemini 2.5 Pro: 32.5 (#213), Gemini 3 Pro: 49.9 (#59)
| Benchmark | Gemini 2.5 Pro | Gemini 3 Pro |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 84.7% | 91.4% |
| Omni-MATH | 41.6% | 55.5% |
| LMArena Math | 1450 | 1476 |
| FrontierMath (Feb 2025 set) | 14.1% | 37.6% |
| FrontierMath Tier 4 (v1) | 4.2% | 18.8% |
| FrontierMath (Tiers 1-3) | 24.6% | — |
| FrontierMath Tier 4 | 0% | — |
| MathArena Final-Answer Competitions | — | 67% |
| ProofBench | — | 20% |
| LiveBench Math | 90.2% | — |
| MATH Level 5 | 95.9% | — |
Knowledge Gemini 3 Pro leads
Gemini 2.5 Pro: 56.0 (#46), Gemini 3 Pro: 64.4 (#16)
| Benchmark | Gemini 2.5 Pro | Gemini 3 Pro |
|---|---|---|
| GPQA Diamond | 85.3% | 92.6% |
| Humanity's Last Exam | 21.6% | 37.5% |
| MMLU-Pro | 86.3% | 90.3% |
| Vectara Hallucination Rate | 7% | 13.6% |
| GPQA (HELM) | 74.9% | 80.3% |
| LMArena Expert | 1452 | 1475 |
| Confabulations | 10.6% | — |
Multimodal Gemini 3 Pro leads
Gemini 2.5 Pro: 45.2 (#18), Gemini 3 Pro: 57.6 (#2)
| Benchmark | Gemini 2.5 Pro | Gemini 3 Pro |
|---|---|---|
| LMArena Vision | 1263 | 1305 |
| GeoBench | 86% | 84% |
| VPCT | 48% | 91% |
| LMArena Document | 1421 | 1434 |
| SpatialViz-Bench | 44.7% | — |
Multilingual Gemini 3 Pro leads
Gemini 2.5 Pro: 55.3 (#31), Gemini 3 Pro: 56.9 (#16)
| Benchmark | Gemini 2.5 Pro | Gemini 3 Pro |
|---|---|---|
| LMArena Non-English | 1451 | 1474 |
| LMArena Chinese | 1507 | 1523 |
| LMArena French | 1472 | 1492 |
| LMArena German | 1487 | 1515 |
| LMArena Japanese | 1461 | 1510 |
| LMArena Korean | 1434 | 1448 |
| LMArena Russian | 1461 | 1493 |
| LMArena Spanish | 1473 | 1470 |
Instruction Following Gemini 3 Pro leads
Gemini 2.5 Pro: 75.0 (#75), Gemini 3 Pro: 76.3 (#45)
| Benchmark | Gemini 2.5 Pro | Gemini 3 Pro |
|---|---|---|
| IFEval | 84% | 87.7% |
| LMArena Instruction Following | 1437 | 1458 |
| LiveBench Instruction Following | 80.6% | — |
Long Context Gemini 2.5 Pro leads
Gemini 2.5 Pro: 59.8 (#5), Gemini 3 Pro: 44.0 (#79)
| Benchmark | Gemini 2.5 Pro | Gemini 3 Pro |
|---|---|---|
| LMArena Longer Query | 1449 | 1471 |
| Fiction.LiveBench | 91.7% | — |
| CL-bench | — | 15.8% |
Writing & Preference Gemini 3 Pro leads
Gemini 2.5 Pro: 63.7 (#62), Gemini 3 Pro: 66.4 (#35)
| Benchmark | Gemini 2.5 Pro | Gemini 3 Pro |
|---|---|---|
| LMArena Text | 1458 | 1479 |
| LMArena Creative Writing | 1454 | 1482 |
| EQ-Bench Creative Writing | 1421 | 1525 |
| WildBench | 85.7% | 85.9% |
| LMArena Multi-Turn | 1453 | 1484 |
| Short-Story Creative Writing | 83.8% | — |
| LiveBench Language | 67.8% | — |
Frequently asked questions
Is Gemini 2.5 Pro better than Gemini 3 Pro?
Gemini 3 Pro is the stronger model overall, scoring 54.8 to 45.0 on the Noometry Index.
Is Gemini 2.5 Pro or Gemini 3 Pro better for coding?
Gemini 3 Pro scores higher on coding benchmarks: 51.6 versus 42.4 in the Noometry coding category.
How many benchmarks do Gemini 2.5 Pro and Gemini 3 Pro share?
58 benchmarks have published results for both models. Gemini 2.5 Pro has 78 scored results on Noometry and Gemini 3 Pro has 67.