Compare
Gemini 2.5 Deep Think vs Gemini 3.1 Pro
Gemini 2.5 Deep Think (Google,Google DeepMind) and Gemini 3.1 Pro (Google DeepMind) compared on benchmarks, pricing, context window, and use-case rankings.
| Attribute | Gemini 2.5 Deep Think | Gemini 3.1 Pro |
|---|---|---|
| Scores | ||
| Intelligence (ECI) | — | 155 |
| Coding | — | 65 |
| Math | 55 | 74 |
| Reasoning & Knowledge | — | 94 |
| Agentic & Tools | — | 83 |
| Specifications | ||
| Developer | Google,Google DeepMind | Google DeepMind |
| Family | Gemini | Gemini |
| Released | Aug 1, 2025 | Feb 19, 2026 |
| Parameters | — | — |
| Availability | Hosted access (no API) | API access |
| Context window | — | — |
| Price — $/M input | — | — |
| Price — $/M output | — | — |
| Inputs | — | — |
| Outputs | — | — |
| Benchmarks | ||
| FrontierMath | 29% | 37% |
| FrontierMath Tier 4 | 10% | 17% |
| AIME 2024/2025 | — | 96% |
| APEX | — | 34% |
| ARC-AGI | — | 98% |
| ARC-AGI-2 | — | 77% |
| GPQA Diamond | — | 94% |
| GSO (code optimization) | — | 23% |
| Humanity's Last Exam | — | 46% |
| METR task horizon | — | 6.4 h |
| SimpleBench | — | 80% |
| SimpleQA Verified | — | 77% |
| SWE-bench Verified | — | 76% |
| Terminal-Bench | — | 80% |
| WebDev Arena | — | 1461 |
| WeirdML | — | 72% |
Use-case scores are 0–100 percentile composites across each area’s benchmarks, ranked against every model from the past year. Highlighted cells lead each row. Open a model for the full picture.
Want a different match-up? Open the compare tool to add or swap models.
More comparisons
Benchmarks & model data from Epoch AI (CC BY); pricing & specs from OpenRouter. ECI = Epoch Capabilities Index.