Compare models
Pick up to 4 models for a side-by-side spec sheet. Hover any dotted metric name for what the test measures. Best value in each row is highlighted.
Claude Opus 5 (batch)GPT-4.1
| Claude Opus 5 (batch) Anthropic | GPT-4.1 OpenAI | |
|---|---|---|
| Pricing | ||
| Input $/1M | $2.50 | $2.00 |
| Output $/1M | $12.50 | $8.00 |
| Blended $/1M (3:1) | $5.00 | $3.50 |
| Cache read $/1M | $0.25 | $0.50 |
| Cache write $/1M | $3.13 | — |
| Quality indices | ||
| Intelligence Index | 50.8 | 12.7 |
| Coding Index | 78 | — |
| Math Index | — | 34.7 |
| Agentic Index | 56.5 | — |
| Speed & latency | ||
| Output speed (tok/s) | 54 | 0 |
| Time to first token | 25.13s | 0s |
| Time to first answer token | 25.13s | 0s |
| Benchmarks (Artificial Analysis) | ||
| MMLU-Pro | — | 80.6% |
| GPQA Diamond | 93.2% | 66.6% |
| Humanity's Last Exam | 54.9% | 4.2% |
| LiveCodeBench | — | 45.7% |
| SciCode | 56.4% | — |
| MATH-500 | — | 91.3% |
| AIME | — | 43.7% |
| AIME 2025 | — | 34.7% |
| IFBench | — | 43.0% |
| AA-LCR (Long Context Reasoning) | 79.3% | 68.3% |
| Terminal-Bench Hard | — | 13.6% |
| Terminal-Bench 2.1 | 89.1% | — |
| τ²-Bench (Telecom) | — | 47.1% |
| τ³-Bench Banking | 42.1% | — |
| Design Arena (Elo) | ||
| Website | 1,321 | 1,051 |
| Web apps | 1,278 | — |
| Full-stack | 1,314 | — |
| Mobile apps | 1,348 | — |
| Android native | 1,266 | — |
| UI components | 1,360 | 1,016 |
| Data viz | 1,353 | 1,117 |
| SVG | 1,345 | — |
| 3D | 1,360 | 878 |
| Game dev | 1,361 | 1,103 |
| Agentic game dev | 1,261 | — |
| Godot game dev | 1,275 | — |
| Python→PPTX slides | 1,265 | — |
| Code categories | 1,338 | 1,041 |
| ASCII art | 1,390 | — |
| Specs | ||
| Context window | 1M | 1.05M |
| Max output tokens | 128K | 33K |
| Input modalities | text, image, file | image, text, file |
| Tokenizer | Claude | GPT |