# MiniMax M3

Sparse MoE with MiniMax Sparse Attention (MSA). 15.6x faster decoding at 1M context. Multimodal (text, image, video).

## Overview

- **Provider:** MiniMax
- **Model ID:** `minimax-m3`
- **Type:** Chat
- **Canonical URL:** https://modelstack.cc/models/chat/minimax-m3
- **Version:** M3
- **Best for:** Frontier coding, long-horizon agentic tasks, multimodal analysis

## Pricing

| Usage | Price |
| --- | ---: |
| Input | $0.3 / 1M tokens |
| Output | $1.2 / 1M tokens |
| Cache read | $0.06 / 1M tokens |
| Cache write (5 minutes) | $0.3 / 1M tokens |

## Specifications

- **Context window:** 1M
- **Maximum output:** 131K
- **Typical latency:** 14 ms
- **Released:** 2026-05-31

## Capabilities

- chat
- tool_use
- coding

## API usage

Use this model ID in a ModelStack API request:

```json
{
  "model": "minimax-m3"
}
```

For the rendered page, visit https://modelstack.cc/models/chat/minimax-m3.