# DeepSeek V3.2

Reasoning-first model built for agents. Competes with GPT-5 level performance.

## Overview

- **Provider:** DeepSeek
- **Model ID:** `deepseek-v3.2`
- **Type:** Chat
- **Canonical URL:** https://modelstack.cc/models/chat/deepseek-v3.2
- **Version:** V3.2
- **Best for:** Reasoning, coding, math, agentic workflows, cost-effective intelligence

## Pricing

| Usage | Price |
| --- | ---: |
| Input | $0.26 / 1M tokens |
| Output | $0.378 / 1M tokens |
| Cache read | $0.13 / 1M tokens |
| Cache write (5 minutes) | $0.26 / 1M tokens |

## Specifications

- **Context window:** 160K
- **Maximum output:** 66K
- **Typical latency:** 14 ms
- **Released:** 2025-12

## Capabilities

- chat
- tool_use
- reasoning
- coding
- analysis

## API usage

Use this model ID in a ModelStack API request:

```json
{
  "model": "deepseek-v3.2"
}
```

For the rendered page, visit https://modelstack.cc/models/chat/deepseek-v3.2.