Qwen3.5 397B A17B pricing & benchmarks
Qwen3.5 397B A17B is available from Alibaba at $0.500 per million input tokens and $3.60 per million output tokens ($1.27 blended at 3:1). It accepts up to 262,144 tokens of context. The Qwen3.5 series 397B-A17B native vision-language model is built on a hybrid architecture that integrates a linear attention mechanism with a sparse mixture-of-experts model, achieving higher inference efficiency. It delivers...
Pricing
Per million tokens · OpenRouter
Input$0.500
Output$3.60
Blended (3:1)$1.27
Specification
qwen/qwen3.5-397b-a17b
Context window262K
Max output236K
ReleasedFeb 2026
LicenceOpen weights
Modalitiesin:text, in:image, in:video, out:text
Benchmarks
Independently run — not vendor self-reports. Every score links its source.
| Benchmark | Score | Config | Run | Source |
|---|---|---|---|---|
| GPQA Diamond | 86.4% ±2.4 | no thinking effort | 2026-08-07 | Independently run · Epoch AI |
| FrontierMath T1–3 | 31.2% ±2.7 | no thinking effort | 2026-08-30 | Independently run · Epoch AI |
| OTIS Mock AIME | 88.9% ±4.7 | thinking effort | 2026-08-07 | Independently run · Epoch AI |
Scores published by Epoch AI, 'AI Benchmarking Hub', used under CC BY 4.0.
Cheaper models that score at least as well on GPQA Diamond
Strictly lower blended price AND an equal-or-higher GPQA Diamond score. One benchmark is one dimension — a model that wins here may still be weaker at your specific task.
| Model | Provider | Blended $/1M | GPQA Diamond | |
|---|---|---|---|---|
| GPT-5.6 Luna | OpenAI | $0.450 (−65%) | 91.6% | Compare → |
| MiniMax M3 | MiniMax | $0.525 (−59%) | 90.9% | Compare → |
| Qwen3.7 Plus | Alibaba | $0.560 (−56%) | 87.9% | Compare → |
| Inkling Small | Thinkingmachines | $0.637 (−50%) | 88.5% | Compare → |
| DeepSeek V4 Flash 0731 | DeepSeek | $0.660 (−48%) | 91.0% | Compare → |
Where to run Qwen3.5 397B A17B
10 hosts · 1.9× between cheapest and dearest. Blended at 3:1. ⚠ marks an option serving less context than the best available, or re-pricing above a prompt-length threshold. Bracketed labels are the provider's own service tier or region.
| Host | Blended $/1M | In / Out | Context | Weights | Uptime 24h |
|---|---|---|---|---|---|
| Alibaba | $0.877 | $0.390 / $2.34 | 262K | — | 99.9% |
| DeepInfra | $1.09 | $0.450 / $3.00 | 262K | fp8 | 99.7% |
| Parasail | $1.27 | $0.500 / $3.60 | 262K | fp8 | 99.6% |
| DigitalOcean ⚠ 131K context, not 262K |
$1.29 | $0.550 / $3.50 | 131K | — | 96.5% |
| Phala | $1.29 | $0.550 / $3.50 | 262K | — | 99.7% |
| AtlasCloud | $1.29 | $0.550 / $3.50 | 262K | fp8 | 100.0% |
| StreamLake | $1.35 | $0.600 / $3.60 | 256K | — | 99.9% |
| GMICloud | $1.35 | $0.600 / $3.60 | 262K | fp8 | 100.0% |
| Novita | $1.35 | $0.600 / $3.60 | 262K | — | 100.0% |
| Venice ⚠ 128K context, not 262K |
$1.69 | $0.750 / $4.50 | 128K | — | 94.5% |
Host prices and uptime from OpenRouter, synced 2026-10-22. Lower-precision weights (fp4, fp8) can cost less and score worse than the same model at bf16 — the score above is not host-specific.
Compare Qwen3.5 397B A17B
FAQ
How much does Qwen3.5 397B A17B cost?
Qwen3.5 397B A17B costs $0.500 per million input tokens and $3.60 per million output tokens on OpenRouter. At a 3:1 input:output mix that blends to $1.27 per million tokens.
What is Qwen3.5 397B A17B's context window?
Qwen3.5 397B A17B accepts up to 262,144 tokens of context, and can generate up to 235,929 output tokens.
How good is Qwen3.5 397B A17B on benchmarks?
Qwen3.5 397B A17B scores 86.4% on GPQA Diamond, independently run and published by Epoch AI. 3 benchmark results are listed on this page, each with its run date and source.
Is there a cheaper model as good as Qwen3.5 397B A17B?
Yes — 5 models in our catalogue cost less per token than Qwen3.5 397B A17B and score at least as high on the same benchmark. The cheapest is GPT-5.6 Luna at $0.450 per million tokens blended.