Qwen vision-language model for visual reasoning, documents, and agent tasks
Benchmarks
| Benchmark | Category | Score | Variant | Provenance | Source | Observed |
|---|---|---|---|---|---|---|
| AA Agentic Index | agentic | 27.0 | — | mirror | openrouter | Jul 23, 2026 |
| 23.3 | non-reasoning | independent | aa | Jul 24, 2026 | ||
| AA Coding Index | coding | 53.7 | — | mirror | openrouter | Jul 23, 2026 |
| 46.6 | non-reasoning | independent | aa | Jul 24, 2026 | ||
| AA Intelligence Index | intelligence | 37.1 | — | mirror | openrouter | Jul 23, 2026 |
| 30.5 | non-reasoning | independent | aa | Jul 24, 2026 | ||
| Median output speed | speed | 55 tok/s | non-reasoning | independent | aa | Jul 24, 2026 |
| Median time to first token | speed | 3.80s | non-reasoning | independent | aa | Jul 24, 2026 |
Includes index data from Artificial Analysis.
Providers
Live operational stats per hosting endpoint, via OpenRouter.
| Provider | Quant | Context | $/1M in · out | Uptime 24h | Latency | Throughput |
|---|---|---|---|---|---|---|
| Alibaba | — | 262K | $0.45 · $2.7 | 99.76% | — | — |
| Chutes | fp8 | 262K | $0.3 · $2 | 96.00% | — | — |
| CoreWeave | fp8 | 262K | $0.6 · $3.6 | 99.89% | — | — |
| DeepInfra | fp8 | 262K | $0.32 · $3.2 | 87.65% | — | — |
| DekaLLM | — | 262K | $0.288 · $2.4 | 98.24% | — | — |
| Io Net | fp8 | 33K | $0.308 · $2.59 | 99.52% | — | — |
| Morph | — | 131K | $0.289 · $2.4 | 97.70% | — | — |
| Phala | — | 262K | $0.32 · $2.7 | 94.39% | — | — |
| SiliconFlow | fp8 | 262K | $0.3 · $3.2 | 93.03% | — | — |
| Venice | fp8 | 256K | $0.325 · $3.25 | 97.24% | — | — |