Compare models
Comparing 2models. Drop the URL into a doc — it's permalinked.
| Field | google/gemini-2.5-pro | deepseek/deepseek-reasoner |
|---|---|---|
| Provider | deepseek | |
| Model ID | gemini-2.5-pro | deepseek-reasoner |
| Context | 1.0M | 64K |
| Max output | 8K | 8K |
| Input / 1M | $1.25 | $0.55 |
| Output / 1M | $10.00 | $2.19 |
| Cached input / 1M | $0.31 | $0.14 |
| Avg cost / 1M | $5.63 | $1.37 |
| Speed | 130 t/s | 30 t/s |
| Quality index | 80.0 | 76.0 |
| MMLU | 89.2 | 90.8 |
| GPQA | 84.0 | 71.5 |
| HumanEval | 92.6 | 86.4 |
| MATH | 91.6 | 97.3 |
| SWE-bench | 63.8 | — |
| Arena Elo | 1380 | — |
| Tools | ✓ | — |
| Vision | ✓ | — |
| Thinking | ✓ | ✓ |
| Streaming | ✓ | ✓ |
| JSON mode | ✓ | — |
| Structured output | ✓ | — |
| Prompt cache | — | — |
Same data, in your terminal: relay models compare gemini-pro r1