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.
GPT-5.6 SolQwen3 235B A22B Instruct 2507
| GPT-5.6 Sol OpenAI | Qwen3 235B A22B Instruct 2507 Alibaba (Qwen) | |
|---|---|---|
| Pricing | ||
| Input $/1M | $2.00 | $0.087 |
| Output $/1M | $10.00 | $0.35 |
| Blended $/1M (3:1) | $4.00 | $0.153 |
| Cache read $/1M | $0.20 | $0.018 |
| Cache write $/1M | $2.50 | — |
| Quality indices | ||
| Intelligence Index | 47 | 12 |
| Coding Index | 77.4 | — |
| Math Index | — | 71.7 |
| Agentic Index | 50.2 | — |
| Speed & latency | ||
| Output speed (tok/s) | 65 | 0 |
| Time to first token | 64.46s | 0s |
| Time to first answer token | 64.46s | 0s |
| Benchmarks (Artificial Analysis) | ||
| MMLU-Pro | — | 82.8% |
| GPQA Diamond | 94.1% | 75.3% |
| Humanity's Last Exam | 49.5% | 11.1% |
| LiveCodeBench | — | 52.4% |
| SciCode | 57.1% | — |
| MATH-500 | — | 98.0% |
| AIME | — | 71.7% |
| AIME 2025 | — | 71.7% |
| IFBench | 72.7% | 46.1% |
| AA-LCR (Long Context Reasoning) | 84.0% | 33.9% |
| Terminal-Bench Hard | 65.9% | 15.2% |
| Terminal-Bench 2.1 | 88.0% | — |
| τ²-Bench (Telecom) | 85.1% | 33.3% |
| τ³-Bench Banking | 44.3% | — |
| Design Arena (Elo) | ||
| Website | — | 1,070 |
| UI components | — | 981 |
| Data viz | — | 1,084 |
| 3D | — | 1,023 |
| Game dev | — | 975 |
| Code categories | — | 1,054 |
| Specs | ||
| Context window | 1.05M | 262K |
| Max output tokens | 128K | 236K |
| Input modalities | file, image, text | text |
| Tokenizer | GPT | Qwen3 |