DeepSeek V3.1 pricing & benchmarks
Released Aug 21, 2025 · served by 8 providers
What DeepSeek V3.1 is
DeepSeek-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...
- Model ID:
deepseek/deepseek-chat-v3.1 - Modality: text->text
- Knowledge cutoff: 2025-03-31
- Tool calling: supported · Structured output: supported
- Weights:
deepseek-ai/DeepSeek-V3.1on Hugging Face
Providers and prices for DeepSeek V3.1
The same model costs different amounts depending on who serves it. Prices are per 1M tokens, cheapest first. Uptime and throughput are OpenRouter's rolling measurements, not vendor claims.
| Provider | Input | Output | Context | Quant | Uptime 24h | Throughput |
|---|---|---|---|---|---|---|
| DeepInfracheapest | $0.250 | $0.950 | 164K | fp4 | 99.6% | — |
| AtlasCloud | $0.300 | $0.950 | 131K | fp8 | 98.8% | — |
| Novita | $0.270 | $1.00 | 131K | fp8 | 100.0% | — |
| SiliconFlow | $0.270 | $1.00 | 164K | fp8 | 93.0% | — |
| SambaNova | $0.650 | $1.50 | 131K | fp8 | 99.6% | — |
| CoreWeave | $0.550 | $1.65 | 161K | fp8 | 94.9% | — |
| $0.600 | $1.70 | 164K | — | 95.7% | — | |
| Mara | $0.600 | $1.70 | 131K | — | 62.5% | — |
DeepSeek V3.1 benchmark results
Design Arena head-to-head results, as reported through the OpenRouter model API.
| Category | Arena | Rank | Elo | Win rate |
|---|---|---|---|---|
| 3d | models | #71 | 1135 | 48% |
| gamedev | models | #71 | 1141 | 47.2% |
| svg | models | #73 | 1017 | 38.2% |
| uicomponent | models | #74 | 1121 | 47.5% |
| website | models | #74 | 1147 | 48% |
| codecategories | models | #75 | 1142 | 47.9% |
| dataviz | models | #75 | 1128 | 46.8% |
FAQ
How much does the DeepSeek V3.1 API cost?
$0.250 per 1M input tokens and $0.950 per 1M output tokens. A typical workload of 1M input + 300K output tokens costs about $0.54.
Is DeepSeek V3.1 free?
No. It is a paid model starting at $0.250 per 1M input tokens, though some providers offer trial credits.
Which provider is cheapest for DeepSeek V3.1?
DeepInfra at $0.250 input / $0.950 output per 1M tokens (fp4 quantization).
Can I self-host DeepSeek V3.1?
Yes — weights are published as deepseek-ai/DeepSeek-V3.1 on Hugging Face, so you can run it on your own hardware.