DeepSeek V3.1
DeepSeekDeepSeek-V3.1 is a large hybrid reasoning model (671B parameters, 37B active) that supports both thinking and non-thinking modes via prompt templates. It extends the DeepSeek-V3 base with a two-phase long-context...
In
$0.25
Out
$0.95
Blended
$0.43
Where to buy this model
8 access routes| Route | In | Out | Blended | vs |
|---|---|---|---|---|
| $0.25 | $0.95 | $0.43 | - | |
| ● SiliconFlow fp8 | $0.27 | $1.00 | $0.45 | +5% |
| $0.27 | $1.00 | $0.45 | +5% | |
| ● AtlasCloud fp8 | $0.30 | $0.95 | $0.46 | +7% |
| ● CoreWeave fp8 | $0.55 | $1.65 | $0.83 | +93% |
| $0.65 | $1.50 | $0.86 | +100% | |
| ● Mara | $0.60 | $1.70 | $0.88 | +105% |
| $0.60 | $1.70 | $0.88 | +105% |
● official API · ● cloud platform · ● inference host · USD per 1M tokens. Same model, different providers - price and quantization vary.
- Ctx
- 164K
- max output
- 33K
- cache read /1M
- $0.13
- Family
- deepseek
- Modalities
- text
- Features
- json · reasoning · tools
Live pricing, updated 2026-09-15.