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 4.6Qwen3 235B A22B Instruct 2507
| Claude Opus 4.6 Anthropic | Qwen3 235B A22B Instruct 2507 Alibaba (Qwen) | |
|---|---|---|
| Pricing | ||
| Input $/1M | $5.00 | $0.09 |
| Output $/1M | $25.00 | $0.55 |
| Blended $/1M (3:1) | $10.00 | $0.205 |
| Cache read $/1M | $0.50 | — |
| Cache write $/1M | $6.25 | — |
| Quality indices | ||
| Intelligence Index | 37.8 | 18.2 |
| Coding Index | — | — |
| Math Index | — | 71.7 |
| Agentic Index | — | — |
| Speed & latency | ||
| Output speed (tok/s) | 54 | 74 |
| Time to first token | 1.89s | 1.12s |
| Time to first answer token | 1.89s | 1.12s |
| Benchmarks (Artificial Analysis) | ||
| MMLU-Pro | — | 82.8% |
| GPQA Diamond | 84.0% | 75.3% |
| Humanity's Last Exam | 18.6% | 10.6% |
| LiveCodeBench | — | 52.4% |
| SciCode | 45.7% | 36.0% |
| MATH-500 | — | 98.0% |
| AIME | — | 71.7% |
| AIME 2025 | — | 71.7% |
| IFBench | 44.6% | 46.1% |
| AA-LCR (Long Context Reasoning) | 58.3% | 31.2% |
| Terminal-Bench Hard | 48.5% | 15.2% |
| τ²-Bench (Telecom) | 84.8% | 33.3% |
| Design Arena (Elo) | ||
| Website | 1,323 | 1,085 |
| Web apps | 1,261 | — |
| Full-stack | 1,279 | — |
| Mobile apps | 1,280 | — |
| Android native | 1,195 | — |
| UI components | 1,336 | 1,009 |
| Data viz | 1,322 | 1,095 |
| SVG | 1,274 | — |
| 3D | 1,335 | 1,053 |
| Game dev | 1,337 | 1,010 |
| Code categories | 1,327 | 1,072 |
| ASCII art | 1,298 | — |
| Specs | ||
| Context window | 1M | 262K |
| Max output tokens | 128K | 16K |
| Input modalities | text, image, file | text |
| Tokenizer | Claude | Qwen3 |