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 4.6Qwen3.8 2.4T A95B (batch)
| Claude Opus 4.6 Anthropic | Qwen3.8 2.4T A95B (batch) Alibaba (Qwen) | |
|---|---|---|
| Pricing | ||
| Input $/1M | $5.00 | $2.00 |
| Output $/1M | $25.00 | $6.00 |
| Blended $/1M (3:1) | $10.00 | $3.00 |
| Cache read $/1M | $0.50 | $0.25 |
| Cache write $/1M | $6.25 | — |
| Quality indices | ||
| Intelligence Index | 26.4 | 39.9 |
| Coding Index | — | 71.9 |
| Math Index | — | — |
| Agentic Index | — | 50.1 |
| Speed & latency | ||
| Output speed (tok/s) | 0 | 39 |
| Time to first token | 0s | 1.88s |
| Time to first answer token | 0s | 53.08s |
| Benchmarks (Artificial Analysis) | ||
| GPQA Diamond | 84.0% | 93.5% |
| Humanity's Last Exam | 19.1% | 42.4% |
| SciCode | — | 54.1% |
| IFBench | 44.6% | — |
| AA-LCR (Long Context Reasoning) | 67.0% | 80.3% |
| Terminal-Bench Hard | 48.5% | — |
| Terminal-Bench 2.1 | — | 82.0% |
| τ²-Bench (Telecom) | 84.8% | — |
| τ³-Bench Banking | — | 49.1% |
| Design Arena (Elo) | ||
| Website | 1,303 | — |
| Web apps | 1,219 | — |
| Full-stack | 1,229 | — |
| Mobile apps | 1,238 | — |
| Android native | 1,216 | — |
| UI components | 1,300 | — |
| Data viz | 1,296 | — |
| SVG | 1,251 | — |
| 3D | 1,303 | — |
| Game dev | 1,301 | — |
| Agentic game dev | 1,215 | — |
| Code categories | 1,304 | — |
| ASCII art | 1,273 | — |
| Specs | ||
| Context window | 1M | 1.01M |
| Max output tokens | 128K | 909K |
| Input modalities | text, image, file | text |
| Tokenizer | Claude | Qwen |