MiniMax

Connect to MiniMax's model API.

API Configuration

API Type	`anthropic-messages`
Base URL	`https://api.minimax.cn/anthropic`
Auth Mode	api_key
Auth Keys	`MINIMAX_API_KEY` — API Key — obtain from https://platform.minimaxi.com/
Regions	cn
Default Model	`MiniMax-M3`
Status	active

Pricing

Pay-as-you-go per token, billed in CNY by model. MiniMax-M3 is tiered by input length: up to 512k input tokens 4.20 input / 16.80 output / 0.84 cache read per million, above 512k 8.40 / 33.60 / 1.68 — MiniMax lists M3 at a permanent 50% discount (2.10 / 8.40 / 0.42, and 4.20 / 16.80 / 0.84 above 512k). MiniMax-M2.7 2.1 / 8.4 / 0.42 cache read and MiniMax-M2.5 2.1 / 8.4 / 0.21, both with cache write 2.625. The priority service tier bills at 1.5x. Verified 2026-09-21 at platform.minimax.cn/docs/guides/pricing-paygo.

Getting Started

Models (3)

Model	Context	Max Output	Pricing
MiniMax-M3 `MiniMax-M3` Newest flagship M-series model: frontier coding and agent capabilities, 1,000,000-token context, native multimodal input (text and images; the API also accepts video, which this catalog does not expose). Thinking is off by default.	1.0M	-	CNY per million tokens, tiered by input length. Up to 512k input: 4.20 input / 16.80 output / 0.84 cache read (permanent 50% discount: 2.10 / 8.40 / 0.42). Above 512k input: 8.40 / 33.60 / 1.68 (discounted 4.20 / 16.80 / 0.84).
MiniMax-M2.7 `MiniMax-M2.7` Historical M-series text model (still served), roughly 60 TPS output; superseded by MiniMax-M3.	205K	-	Input $0.30/M, output $1.20/M (same as M2.5).
MiniMax-M2.5 `MiniMax-M2.5` SOTA model for productivity, coding (80.2% SWE-Bench), and agent workflows	205K	-	Input $0.30/M, output $1.20/M. Cache read $0.03/M, cache write $0.375/M.

Last verified: 2026-09-21