# deepseek-v3.2-exp — API, Pricing & Context Window

deepseek-v3.2-exp on Vivgrid: DeepSeek's experimental V3.2 model with a 164K context window, available through one unified API.

`deepseek-v3.2-exp` is the experimental branch of DeepSeek V3.2, offering an extended **164K-token context window** and a testbed for the latest DeepSeek improvements. It's intended for teams that want to evaluate cutting-edge behavior before it lands in stable releases.

Vivgrid exposes `deepseek-v3.2-exp` through the same unified, OpenAI-compatible API as the rest of the catalog, so experimentation requires no extra integration work.

## Specifications

|     |     |
| --- | --- |
| **Provider** | DeepSeek |
| **Model ID** | `deepseek-v3.2-exp` |
| **Best for** | General-purpose |
| **Context window** | 163,840 tokens |
| **Modalities** | Text |
| **Tool / function calling** | Yes |
| **Knowledge cutoff** | 2025-01 |
| **Acceleration** | 🌐 Global (Centralized) |

## Pricing

Pricing for `deepseek-v3.2-exp` matches the provider's published rates. See live pricing in the [Vivgrid Console](https://console.vivgrid.com/).

## Quick start

Call `deepseek-v3.2-exp` through Vivgrid's unified, OpenAI-compatible endpoint. Get an API key from the [Vivgrid Console](https://console.vivgrid.com/).

```
curl https://api.vivgrid.com/v1/chat/completions \
  -H "Authorization: Bearer $VIVGRID_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v3.2-exp",
    "messages": [
      { "role": "user", "content": "Say hello in English, Chinese and Spanish." }
    ],
    "stream": true
  }'
```

## [Ideal use cases](/content/docs/models/deepseek-v3.2-exp#ideal-use-cases/index.html)

- Evaluating experimental DeepSeek capabilities
- Longer-context tasks than stable V3.2 (up to 164K tokens)
- A/B testing against `deepseek-v3.2` in production-like setups
- Research and benchmarking workflows

## [Related models](/content/docs/models/deepseek-v3.2-exp#related-models/index.html)

- [deepseek-v3.2](/content/docs/models/deepseek-v3.2) — the stable counterpart
- [deepseek-v4-pro](/content/docs/models/deepseek-v4-pro/index.html) — the flagship next generation
- [deepseek-r1-0528](/content/docs/models/deepseek-r1-0528/index.html) — reasoning-focused model