# MiniMax M3

> MiniMax M3 pairs a 1M-token context window with an unusually large 512K maximum output — the widest output headroom of any model in this catalog. It accepts text and image input, supports tool calling, reasoning and prompt caching, and answers on three protocols: OpenAI Chat Completions, OpenAI Responses and Anthropic Messages.

Canonical: https://routemux.com/minimax/minimax-m3
Provider: MiniMax

## Key facts

| Field | Value |
| --- | --- |
| Model ID | minimax/minimax-m3 |
| Official model ID | MiniMax-M3 |
| Context window | 1M |
| Max output | 512K |
| Input modalities | text, image |
| Output modalities | text |
| Capabilities | Vision, Reasoning, Tool Calling, Coding, Cache |
| Availability | available |

## Pricing

- RouteMux input: $0.03 per 1M tokens
- RouteMux output: $0.12 per 1M tokens
- Official input: $0.3 per 1M tokens
- Official output: $1.2 per 1M tokens
- Cache read: $0.006 per 1M tokens
- Saving vs official: 90%
- Official price source: minimax_text_x0.10, verified 2026-08-07

Billing is prepaid wallet, charged per successful request only.

## API access

Base URL: https://api.routemux.com

Supported protocols:

- openai_chat
- openai_responses
- anthropic_messages

## What is MiniMax M3

MiniMax M3 pairs a 1M-token context window with an unusually large 512K maximum output — the widest output headroom of any model in this catalog. It accepts text and image input, supports tool calling, reasoning and prompt caching, and answers on three protocols: OpenAI Chat Completions, OpenAI Responses and Anthropic Messages.

Through RouteMux it is $0.03 per 1M input tokens against MiniMax's $0.30, and $0.12 against $1.20 on output.

The 512K output ceiling is the thing worth designing around. Tasks that normally have to be split across calls because the response will not fit — full-file rewrites, long structured extractions, bulk translation — can often run in one request here.

## Key capabilities

- **512K maximum output** — The widest output headroom in this catalog. Work that normally needs splitting across calls can run in one.
- **1M-token context** — With text and image input.
- **Tool calling and caching** — Reasoning, tool calling and prompt caching are all supported.
- **Three protocols** — OpenAI Chat, OpenAI Responses and Anthropic Messages.

## Sources

- [MiniMax — API docs](https://platform.minimaxi.com/document)

_Content reviewed by Jerry Fan on 2026-08-09._