# DeepSeek V3.1

> DeepSeek V3.1 is a hybrid reasoning model from DeepSeek with a 131,072-token context window, strong coding and agentic tool-use performance. Served free through SambaNova Cloud's fast RDU inference and SiliconFlow.


**ID**: `deepseek/deepseek-v3.1`  
**Creator**: DeepSeek  
**Category**: text  
**Context**: 131K tokens  
**Pricing**: $0.25 / $0.95 per 1M · billed by your provider (BYOK)  
**Released**: 2025-08-21  
**Web page**: https://anyrouter.dev/model/deepseek/deepseek-v3.1

**Input modalities**: text  
**Output modalities**: text  

**Capabilities**: chat, reasoning, streaming, coding, function-calling

**Aliases**: `deepseek/DeepSeek-V3.1`, `deepseek-ai/DeepSeek-V3.1`

## Usage

**Endpoint**: `POST https://anyrouter.dev/api/v1/chat/completions`

```bash
curl https://anyrouter.dev/api/v1/chat/completions \
  -H "Authorization: Bearer $ANYROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "deepseek/deepseek-v3.1",
  "messages": [
    {
      "role": "user",
      "content": "Say hi in 3 words."
    }
  ]
}'
```

## Providers

| Provider | Input | Output |
| --- | --- | --- |
| sambanova-byok | $0.25/M (list) | $0.95/M (list) |
| siliconflow-byok | $0.25/M (list) | $0.95/M (list) |
| cline-byok | $0.25/M (list) | $0.95/M (list) |
| vercel-byok | $0.25/M (list) | $0.95/M (list) |
| deepinfra-byok | $0.25/M (list) | $0.95/M (list) |
| huggingface-byok | $0.25/M (list) | $0.95/M (list) |

BYOK routes are billed by your own provider key — AnyRouter charges $0. Rates marked (list) are the provider's published list price.

## Benchmarks

- **GPQA Diamond**: 0.8
- **HLE**: 0.1
- **τ²-Bench Telecom**: 0.4
- **SciCode**: 0.4

## Model card

- Architecture: mixture of experts [1]
- Total parameters: 671B [1]
- Active parameters: 37B [1]
- Native context: 131072 tokens [1,2]
- License: MIT, open weights (https://huggingface.co/deepseek-ai/DeepSeek-V3.1) [1]

### Strengths [1,2]

- One model with thinking and non-thinking modes
- Better tool calling and agent tasks than V3
- Thinking mode matches R1-0528 quality and answers faster

### Use cases [1,2]

- Code agents and search agents
- Tool-calling workflows
- Chat and reasoning from one deployment

### API notes [1]

- Thinking mode is switched through the chat template
- Tool calls are supported in non-thinking mode

### Links

- announcement: https://api-docs.deepseek.com/news/news250821
- model card: https://huggingface.co/deepseek-ai/DeepSeek-V3.1
- Previous version: https://anyrouter.dev/model/deepseek/deepseek-v3
- Next version: https://anyrouter.dev/model/deepseek/deepseek-v3.2

### Sources

1. [deepseek-ai/DeepSeek-V3.1 model card](https://huggingface.co/deepseek-ai/DeepSeek-V3.1/raw/main/README.md) (accessed 2026-10-14)
2. [DeepSeek V3.1 Release | DeepSeek API Docs](https://api-docs.deepseek.com/news/news250821) (accessed 2026-10-14)

## Source

- Upstream docs: https://api-docs.deepseek.com