Qwen: Qwen3.5 397B A17B

qwen/qwen3.5-397b-a17b

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...

Prompt $/1M

$0.390

Completion $/1M

$2.340

Context

262,144 tok

Best provider

158 tok/s · Together

Specs

Model id: qwen/qwen3.5-397b-a17b
Display name: Qwen: Qwen3.5 397B A17B
Modality: text+image+video->text
Tokenizer: Qwen3
Instruct type: —
Context length: 262,144
Max completion: 65,536
Prompt $/1M: $0.390
Completion $/1M: $2.340
Free tier: no

Capabilities

Function callingno
JSON modeno
System promptno
Visionyes
Moderatedno

Benchmark scores

Each axis is one benchmark, normalized to the best model in the field.

No benchmark data for this model.

Speed by provider

Provider	tok/s	TTFT	Total latency
Togetherfastest	158	616 ms	888 ms
Wafer	99	1,221 ms	1,363 ms
Nebius	67	879 ms	1,145 ms
Artificial Analysis	65	1,753 ms	—
Morph	54	1,469 ms	1,683 ms
AtlasCloudfastest	54	1,266 ms	1,730 ms
GMICloud	48	2,636 ms	3,208 ms
Parasail	44	334 ms	357 ms
DeepInfra	30	383 ms	462 ms
Venice	29	590 ms	650 ms
StreamLake	25	766 ms	908 ms
Phala	17	3,538 ms	4,203 ms
Novita	17	859 ms	912 ms
Alibaba	16	1,226 ms	1,385 ms
Chutes	12	1,493 ms	1,744 ms
DigitalOcean	7	749 ms	835 ms