How much does the MiniMax API cost?

MiniMax-M3.1 Flash costs $0.15 per 1M input tokens, $0.015 cached and $0.60 output. MiniMax-M3 costs $0.30 in and $1.20 out for prompts up to 512K tokens, as of 2026-11-02.

What is the cheapest MiniMax model?

MiniMax-M3.1 Flash, at $0.15 input and $0.60 output per 1M tokens, with cache reads at $0.015.

What are MiniMax's API rate limits?

MiniMax-M3 and MiniMax-M3.1 Flash both allow 200 requests per minute and 10 million tokens per minute. MiniMax raises limits on request through api@minimax.io.

Why did my MiniMax-M3 request cost double?

Your prompt was probably over 512K input tokens. Above that, MiniMax bills the whole request at $0.60 in and $2.40 out per 1M.

What is MiniMax Priority?

A service tier for MiniMax-M3 that admits your requests first for faster, more reliable responses. It costs 1.5x the Standard price; set service_tier to priority.

Is the M Plan cheaper than the MiniMax API?

It can be for one person coding every day, since it's a flat subscription with 5-hour and weekly usage windows. For apps, bursty traffic or anything in production, the pay-as-you-go API is the right fit.

MiniMax API pricing: M3, M3.1 Flash, rate limits, M Plan

USD per 1M tokens, MiniMax Open Platform pay-as-you-go price, checked 2026-11-02. MiniMax lists M3 at $0.60 in and $2.40 out with a "permanent 50% off", so the table shows what you actually pay. Cached means a prompt-cache read.

Model and tier	Input	Cached	Output	Notes
MiniMax-M3.1 Flash	$0.15	$0.015	$0.60	Newest, cheapest
MiniMax-M3, up to 512K	$0.30	$0.06	$1.20	Standard tier, after the permanent 50% off
MiniMax-M3, over 512K	$0.60	$0.12	$2.40	Whole request at the long-prompt rate
MiniMax-M3 Priority, up to 512K	$0.45	$0.09	$1.80	1.5x Standard, faster admission

Which MiniMax model should you use?

MiniMax-M3.1 Flash for most new work. MiniMax-M3.1 Flash is half the price of M3 on both input and output ($0.15 / $0.60), and number 13 of 77 on our full price list. On LMArena it's #26 of 50 priced models on Code Arena and not ranked yet on Text Arena. It's also the only text model in MiniMax's M Plan subscription.
MiniMax-M3 if you need the Priority tier or want to self-host. MiniMax-M3 is #34 of 50 priced models on Code Arena and #41 of 51 priced models on Text Arena, so on LMArena's numbers it doesn't beat the cheaper Flash at coding. MiniMax publishes its weights, with an official self-hosting guide, so you can run it on your own servers instead of paying per token. That means server hardware: MiniMax's reference setup is eight Nvidia B200 GPUs, and it still marks the recipe experimental.

If you used MiniMax-M3.1-Flash-Preview, that name is retired and MiniMax now serves those requests with MiniMax-M3.1-Flash, per its M Plan docs.

How the MiniMax-M3 512K price step works

MiniMax prices M3 by how long your prompt is. Up to 512K input tokens you pay $0.30 in and $1.20 out. Above 512K, the whole request moves to $0.60 in and $2.40 out, not just the tokens past the line. A single 600K-token prompt with a 10K reply costs about $0.38 by our math, against $0.16 for a 500K prompt. Trim long prompts under 512K when you can.

Standard vs Priority tier on MiniMax-M3

Priority gets your requests admitted first, for faster responses and better reliability when MiniMax is busy. It costs 1.5x Standard: $0.45 in and $1.80 out up to 512K. You turn it on per request by setting service_tier to priority. MiniMax lists Priority for M3 only, not M3.1 Flash, per its pricing page.

How MiniMax prompt caching is priced

A cache read costs 80% less than fresh input on M3 ($0.06) and 90% less on M3.1 Flash ($0.015). MiniMax doesn't list a separate cache-write fee on the pricing page. If you resend the same long system prompt or codebase on every call, caching is the biggest saving you can get.

MiniMax API rate limits

MiniMax-M3 and MiniMax-M3.1 Flash share one published limit: 200 requests per minute and 10,000,000 tokens per minute (input plus output). There are no spend tiers like OpenAI or StepFun have. For more, MiniMax says to email its business team at [email protected], per its rate limits page.

Web search and other tool fees

The server-side web_search tool costs $0.01 per request on top of tokens, and the API-vlm image understanding tool is also $0.01 per call, per MiniMax.

MiniMax M Plan vs the pay-as-you-go API

M Plan is MiniMax's subscription for individuals, with three tiers: Go, Explore at 3x Go's usage and Build at 7.5x. All three use M3.1 Flash for text, and Explore and Build add the H3 video model. Usage resets in two windows, every 5 hours (text, image and audio) and every 7 days (everything, video included), and unused quota doesn't carry over, per MiniMax's usage rules. Annual billing charges 10 months for 12. You use it in MiniMax Code by signing in, or in Claude Code, Codex, Cursor and OpenCode with an M Plan API key. That key is separate from a regular API key, so production apps should stay on pay-as-you-go. MiniMax's docs don't print M Plan prices; check the checkout page before you buy.

What a month costs

List price, no discounts, our math. Medium is 20M input and 5M output tokens a month; heavy is 200M and 50M.

Model	Medium month	Heavy month
MiniMax-M3.1 Flash	$6.00	$60.00
MiniMax-M3	$12.00	$120

MiniMax against the models people compare it with

MiniMax-M3.1 Flash vs DeepSeek V4.1 Flash: $0.26 vs $0.52 blended.
MiniMax-M3.1 Flash vs Gemini 3.8 Flash: $0.26 vs $1.50 blended.
MiniMax-M3 vs DeepSeek V4.1 Flash, MiniMax-M3 vs Claude Sonnet 5.5 and GLM-5.3 vs MiniMax-M3.
All Chinese labs side by side: best Chinese LLM APIs.

MiniMax API pricing explained