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 4.5 (batch)GPT-5.6 Terra
| Claude Opus 4.5 (batch) Anthropic | GPT-5.6 Terra OpenAI | |
|---|---|---|
| Pricing | ||
| Input $/1M | $2.50 | $2.00 |
| Output $/1M | $12.50 | $12.00 |
| Blended $/1M (3:1) | $5.00 | $4.50 |
| Cache read $/1M | $0.25 | $0.20 |
| Cache write $/1M | $3.13 | $2.50 |
| Quality indices | ||
| Intelligence Index | 23.7 | 42.1 |
| Coding Index | — | 76.7 |
| Math Index | 62.7 | — |
| Agentic Index | — | 43.2 |
| Speed & latency | ||
| Output speed (tok/s) | 0 | 88 |
| Time to first token | 0s | 130.06s |
| Time to first answer token | 0s | 130.06s |
| Benchmarks (Artificial Analysis) | ||
| MMLU-Pro | 88.9% | — |
| GPQA Diamond | 81.0% | 92.5% |
| Humanity's Last Exam | 13.2% | 42.9% |
| LiveCodeBench | 73.8% | — |
| SciCode | — | 55.0% |
| AIME 2025 | 62.7% | — |
| IFBench | 43.0% | 71.2% |
| AA-LCR (Long Context Reasoning) | 70.7% | 83.0% |
| Terminal-Bench Hard | 40.9% | 57.6% |
| Terminal-Bench 2.1 | — | 88.0% |
| τ²-Bench (Telecom) | 86.3% | 86.3% |
| τ³-Bench Banking | — | 40.2% |
| Design Arena (Elo) | ||
| Website | 1,259 | — |
| Web apps | 1,174 | — |
| Full-stack | 1,166 | — |
| Mobile apps | 1,218 | — |
| Android native | 1,190 | — |
| UI components | 1,254 | — |
| Data viz | 1,258 | — |
| SVG | 1,204 | — |
| 3D | 1,242 | — |
| Game dev | 1,252 | — |
| Code categories | 1,257 | — |
| ASCII art | 1,208 | — |
| Specs | ||
| Context window | 200K | 1.05M |
| Max output tokens | 64K | 128K |
| Input modalities | file, image, text | file, image, text |
| Tokenizer | Claude | GPT |