Qwen3.5 397B A17B
- Input / output
- $0.36 / $2.40
- Providers
- 8
- Spread
- 1.93×
- 7 days
- −0.9%
- 30 days
- −0.9%
- Context
- 256K
Providers
| Provider | Input | Output | Blended | vs ref | Context | Uptime 1d | Price since |
|---|---|---|---|---|---|---|---|
| fp8 | $0.36 | $2.40 | $0.87 | ref | |||
| $0.39 | $2.34 | $0.878 | ref | ||||
| fp8 | $0.45 | $3.00 | $1.09 | +24% | |||
| fp8 | $0.50 | $3.60 | $1.275 | +46% | |||
| fp8 | $0.55 | $3.50 | $1.29 | +47% | |||
| fp8 | $0.60 | $3.60 | $1.35 | +55% | |||
| $0.60 | $3.60 | $1.35 | +55% | ||||
| $0.60 | $3.60 | $1.35 | +55% | ||||
DeepInfrafp8/priority fp8 | $0.675 | $4.50 | $1.63 | +87% | |||
| $0.75 | $4.50 | $1.69 | +93% |
- DeepInfraCheapestfp8/flex$0.87$0.36 in · $2.40 outref vs ref256K ctx100% uptimefp8
- $0.878$0.39 in · $2.34 outref vs ref256K ctx100% uptime
- $1.09$0.45 in · $3.00 out+24% vs ref256K ctx97.8% uptimefp8
- $1.275$0.50 in · $3.60 out+46% vs ref256K ctx99.0% uptimefp8
- $1.29$0.55 in · $3.50 out+47% vs ref256K ctx100% uptimefp8
- $1.35$0.60 in · $3.60 out+55% vs ref256K ctx100% uptimefp8
- $1.35$0.60 in · $3.60 out+55% vs ref256K ctx100% uptime
- $1.35$0.60 in · $3.60 out+55% vs ref250K ctx97.8% uptime
- fp8/priority$1.63$0.675 in · $4.50 out+87% vs ref256K ctx92.5% uptimefp8
- $1.69$0.75 in · $4.50 out+93% vs ref125K ctx98.3% uptime
Blended = (3 × input + output) / 4. Ref $0.874 = median of cheapest 20%.
Price history
USD per 1M tokens · each step is a price change
Prices in USD / 1M tokens.
No provider has changed its price since tracking began on Oct 3, 2026.
Other pricing
| Offered by | Cache read |
|---|---|
| DeepInfra (fp8/flex) | $0.176 |
| DeepInfra | $0.22 |
| Parasail | $0.30 |
| AtlasCloud | $0.55 |
| StreamLake | $0.12 |
| DeepInfra (fp8/priority) | $0.33 |
Changes
- ListedNow served,Listed
- ListedNow served,Listed
- DelistedNo longer served,Delisted
- DelistedNo longer served,Delisted
About
The cheapest provider for Qwen3.5 397B A17B is DeepInfra at $0.87 per 1M tokens (blended), charging $0.36 per 1M input and $2.40 per 1M output tokens. 8 providers serve it; the most expensive charges 1.93× the typical price. Its best price fell 0.9% over the last 30 days. Context window: 256K tokens. Prices tracked since Oct 3, 2026.
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...
- Released
- Feb 16, 2026
- First tracked
- Oct 3, 2026
- Canonical slug
- qwen/qwen3.5-397b-a17b-20260216
- Weights
- Qwen/Qwen3.5-397B-A17B