Side-by-side comparison of AI model specs, benchmarks, and pricing.
| Spec | Claude Haiku 4.5 Anthropic | Gemini 2.5 Pro Google |
|---|---|---|
| Tier | mid | frontier |
| Release Date | 2025-10-01 | 2025-06-05 |
| Context Window | 200K | 1049K |
| Max Output | 64K | 66K |
| Input Price (/1M) | $1 | $1.25 |
| Output Price (/1M) | $5 | $10 |
| Arena Elo | 1,404 | 1,450 |
| MMLU | 82% | 89% |
| GPQA | 55% | 74% |
| MATH | — | 90.2% |
| HumanEval | 90% | 93.2% |
| SWE-bench | — | 63.8% |
| Capabilities | vision, tool-use, code, extended-thinking | vision, tool-use, code, reasoning, audio |