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 3 HaikuGPT-4.1
| Claude 3 Haiku Anthropic | GPT-4.1 OpenAI | |
|---|---|---|
| Pricing | ||
| Input $/1M | $0.25 | $2.00 |
| Output $/1M | $1.25 | $8.00 |
| Blended $/1M (3:1) | $0.50 | $3.50 |
| Cache read $/1M | $0.03 | $0.50 |
| Cache write $/1M | $0.30 | — |
| Quality indices | ||
| Intelligence Index | 3.9 | 19.4 |
| Coding Index | — | — |
| Math Index | — | 34.7 |
| Agentic Index | — | — |
| Speed & latency | ||
| Output speed (tok/s) | 0 | 151 |
| Time to first token | 0s | 0.56s |
| Time to first answer token | 0s | 0.56s |
| Benchmarks (Artificial Analysis) | ||
| MMLU-Pro | — | 80.6% |
| GPQA Diamond | 37.4% | 66.6% |
| Humanity's Last Exam | 3.9% | 4.6% |
| LiveCodeBench | 15.4% | 45.7% |
| SciCode | 18.6% | 38.1% |
| MATH-500 | 39.4% | 91.3% |
| AIME | 1.0% | 43.7% |
| AIME 2025 | — | 34.7% |
| IFBench | 36.1% | 43.0% |
| AA-LCR (Long Context Reasoning) | 21.0% | 61.0% |
| Terminal-Bench Hard | 0.8% | 13.6% |
| τ²-Bench (Telecom) | 21.1% | 47.1% |
| Design Arena (Elo) | ||
| Website | — | 1,066 |
| UI components | — | 1,043 |
| Data viz | — | 1,140 |
| 3D | — | 908 |
| Game dev | — | 1,138 |
| Code categories | — | 1,059 |
| Specs | ||
| Context window | 200K | 1.05M |
| Max output tokens | 4K | 33K |
| Input modalities | text, image | image, text, file |
| Tokenizer | Claude | GPT |