Gemini 2.5 Pro
Google DeepMindGemini 2.5 Pro was released March 25, 2025 as Google's most capable model with built-in reasoning (thinking). It topped the LMSys Chatbot Arena leaderboard on release. Scored 63.8% on SWE-bench Verified, 84% on GPQA Diamond, and 97.3% on MATH. Featured a 1M token context window with reasoning capabilities. Priced at \$1.25/\$10 per million tokens.
Pricing
| Type | Price per MTok |
|---|---|
| Input | $1.25 |
| Output | $10 |
Benchmarks3
| Benchmark | Score |
|---|---|
| SWE-bench Verified | 63.8% |
| GPQA Diamond | 84% |
| MATH | 97.3% |
Gemini Family5
| Model | Tier | Released | Input $/MTok |
|---|---|---|---|
| Gemini 2.5 Flash | flash | 2025-05-20 | $0.15 |
| Gemini 2.0 Flash | flash | 2024-12-11 | $0.1 |
| Gemini 1.5 Flash | flash | 2024-05-24 | $0.075 |
| Gemini 1.5 Pro | pro | 2024-02-15 | $1.25 |
| Gemini 1.0 Ultra | ultra | 2024-02-08 |
Details
Model FamilyGemini
Tierpro
Generation2.5
Release Date2025-03-25
Context Window1M tokens
Capabilities5
reasoningtool-usevisionaudio-inputlong-context
Sources1
Tags
geminigooglereasoningfrontier-ai