OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Benchmark | Score | What it measures | Source | Retrieved |
|---|---|---|---|---|
| Humanity's Last Exam | 18.8% | Expert-level questions across many domains | Source | September 22, 2026 |
| GPQA Diamond | 84.0% | Google-proof graduate science questions (hardest subset) | Source | September 22, 2026 |
| AIME 2025 | 86.7% | Competition mathematics (AIME), 2025 | Source | September 22, 2026 |
| SWE-bench Verified | 63.8% | Resolving verified real GitHub issues | Source | September 22, 2026 |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
Gemini Pro 2.5 Experimental This page | ||||
| Reasoning + agents | ||||
| Balanced coding + agents | ||||
| Long-context, multimodal & agentic workflows | ||||
| Reasoning + agents |
Yes, Gemini Pro 2.5 Experimental can stream responses as they are generated.
Yes, Gemini Pro 2.5 Experimental supports both function calling and structured outputs.
Gemini Pro 2.5 Experimental was built by Google.
It is available on the v1 endpoint.
