Gemini 2.5 Pro
Gemini 2.5 model with thinking built in. Experimental on 25 March 2025, generally available from 17 June.
- Announced
- Released
- Access
- Public
Benchmark scores
| Test | Score | Best on the tracker | How it was measured |
|---|---|---|---|
| Epoch Capabilities Indexmany tests combined into one score | 145 points | GPT-6.2 Astra 174 points | Independentlisted as "Gemini 2.5 Pro (Jun 2025)", CI 143.36 to 146.22Run 9 Nov 2026 |
| Remote Labor Indexpaid freelance projects | 0.8% | Claude Fable 5.2 41% | Independentgemini-2.5-pro-preview-06-05Run 27 Oct 2025 |
| Humanity's Last Examexpert questions, no tools | 22% | GPT-6 Astra 55% | Independentleaderboard entry "gemini-2.5-pro-preview-06-05" (gemini-2.5-pro-preview-06-05) |
| FrontierMath T1–3hard maths problems | 25% | Claude Fable 5.2 95% | Independentdefault settings (gemini-2.5-pro)Run 11 Jun 2026 |
| GPQA DiamondPhD-level science questionsPhD experts: 70%Saturated | 85% | Gemini 4 Argon 96% | Independentdefault settings (gemini-2.5-pro)Run 16 Nov 2025 |
| FrontierMath T4research-level maths problemsSaturated | 0% | GPT-6.1 Sol 100% | Independentdefault settings (gemini-2.5-pro)Run 11 Jun 2026 |
| ARC-AGI-2visual puzzles, model aloneAverage person: 60%Saturated | 4.9% | GPT-6 Astra 95% | IndependentARC Prize config "Gemini 2.5 Pro (Thinking 32K)" (gemini-2-5-pro-2025-06-17-thinking-32k) |
| OTIS Mock AIMEcompetition mathsSaturated | 85% | GPT-5.5 100% | Independentdefault settings (gemini-2.5-pro)Run 16 Nov 2025 |
| SWE-bench Verifiedfixing real software bugsKnown flaws | 58% | Claude Opus 4.7 83% | Independentdefault settings (gemini-2.5-pro)Run 13 Feb 2026 |
Also on these charts
Sources
- Gemini 2.5: Our newest Gemini model with thinkingGoogle · Official · blog.google
- Gemini API release notesGoogle · Official · ai.google.dev
- Epoch AI Benchmarking Hub (model release dates and runs)Epoch AI · Dataset · 9 Nov 2026 · epoch.ai