AI Rate CardQwen3 VL 30B A3B Thinking vs Qwen3.5 397B-A17B: API price and specs
Qwen: Qwen3 VL 30B A3B Thinking is 37% cheaper than Qwen3.5 397B-A17B on input price.
Qwen3.5 397B-A17B is 0% cheaper than Qwen: Qwen3 VL 30B A3B Thinking on output price.
Qwen: Qwen3 VL 30B A3B Thinking is 17% cheaper than Qwen3.5 397B-A17B on a 1M-in / 200k-out workload.
| Field | Qwen: Qwen3 VL 30B A3B Thinking | Qwen3.5 397B-A17B |
|---|---|---|
| Lowest input, per 1M | $0.10 | $0.16 +59.9% |
| Median input, per 1M | $0.29 | $0.55 +89.7% |
| Lowest output, per 1M | $0.99 +0.4% | $0.99 |
| Median output, per 1M | $1.00 | $3.60 +260% |
| Lowest unit price | — | — |
| Cost of 1M in + 200k out | $0.30 | $0.36 +20.2% |
| Context | 262K | 262K |
| Max output | 262K +0.1% | 262K |
| Providers | 6 +80% | 30 |
| Regions | — | CH, EU, GLOBAL, US |
| Accepts | image, image/bmp, image/heic, image/jpeg, image/png, image/tiff, image/webp, text, video/mp4, video/quicktime, video/x-flv, video/x-matroska, video/x-ms-wmv, video/x-msvideo | file, image, image/bmp, image/heic, image/jpeg, image/png, image/tiff, image/webp, text, video, video/mp4, video/quicktime, video/x-flv, video/x-matroska, video/x-ms-wmv, video/x-msvideo |
| Returns | text | text |
| Type | Text models | Text models |
| Free option | no | no |
| Cheapest provider | aihubmix | aihubmix |
Your own workload: how many tokens the model gets (input) and gives back (output) - for example, in a month. The "Cost" row of the table above is recalculated for these numbers, for each model.
Computed from current offers each time the page opens; the best value of a row is bold, the percent is how far behind the best.
Providers selling both
aihubmix, deepinfra, kilo, openrouter, qiniu-ai, siliconflow
Frequently asked questions
Which is cheaper, Qwen: Qwen3 VL 30B A3B Thinking or Qwen3.5 397B-A17B?
Qwen: Qwen3 VL 30B A3B Thinking is 37% cheaper than Qwen3.5 397B-A17B on input price. Qwen3.5 397B-A17B is 0% cheaper than Qwen: Qwen3 VL 30B A3B Thinking on output price. Qwen: Qwen3 VL 30B A3B Thinking is 17% cheaper than Qwen3.5 397B-A17B on a 1M-in / 200k-out workload.
Which has the larger context, Qwen: Qwen3 VL 30B A3B Thinking or Qwen3.5 397B-A17B?
Qwen: Qwen3 VL 30B A3B Thinking: 262K tokens; Qwen3.5 397B-A17B: 262K tokens.
Where can I use both Qwen: Qwen3 VL 30B A3B Thinking and Qwen3.5 397B-A17B?
aihubmix, deepinfra, kilo, openrouter, qiniu-ai, siliconflow