MiniMax M2 Pricing (2026)

MiniMax-M2 is MiniMax's default general-and-coding tier, and the cheapest way in is OpenRouter at $0.255 input and $1.00 output per million tokens, a touch below the $0.30 / $1.20 official MiniMax rate. That small gap is the whole decision for most teams, so this piece walks the routed price, the official price, and what one real call actually cost.

If you are choosing a MiniMax model and do not have a specific reason to pick something else, M2 is the one to default to. It is priced as the workhorse tier, it answered both a general and a coding prompt for us on the first try, and the per-call cost landed in the fractions-of-a-cent range. The question that matters is where you buy it, because the routed price and the official price are not the same number.

MiniMax-M2 API pricing (verified 2026-06)

MiniMax-M2 is the mid-tier model that MiniMax positions as its general-purpose and coding default, sitting below the higher-context M3 and above the legacy M1. We reached it through OpenRouter because we do not hold a native MiniMax key, so every routed figure below is the OpenRouter-billed value, and the official numbers are sourced and still pending a native-key re-check.

The two prices you will see quoted are close but not identical:

Source	Input ($/1M)	Output ($/1M)	Notes
OpenRouter-routed (measured)	$0.255	$1.00	live catalog, billed to our account
Official MiniMax	$0.30	$1.20	needs native re-verify

The routed price is about 15% cheaper on input and 17% cheaper on output, which for a high-volume workload is real money. The practical upshot is that if your traffic already flows through an aggregator, M2 costs you slightly less than going direct, and you inherit the aggregator's billing and failover. According to MiniMax, the official endpoint is the path you would take for native account features, region settings, and any enterprise terms, so the choice is not purely about the headline rate.

According to MiniMax, M2 is offered as the standard tier on its international platform, which is why we treat it as the sensible default rather than a specialized pick. One caution carried across this whole cluster: do not anchor a long-context argument on the older MiniMax-M1, because the published context window for M1 conflicts between sources and should be confirmed in the console before you design around it.

What one MiniMax-M2 call actually cost (first-hand, via OpenRouter)

Documentation gives you a rate card; it does not tell you what a single request feels like. So we sent two prompts to minimax/minimax-m2 through OpenRouter and recorded the billed cost and latency from the response itself.

Prompt type	Prompt tokens	Output tokens	Billed cost (USD)	Latency
General	39	200	$0.000209945	5.08s
Coding	38	200	$0.0002514	2.95s

A few things stood out. The coding call came back in 2.95 seconds, noticeably faster than the 5.08-second general call, even though it billed slightly more because it used a hair more of the priced output budget. Both completions hit the 200-token output cap we set, so treat 200 as capped rather than the model's natural stopping point. And both costs are tiny in absolute terms: a fifth of a thousandth of a dollar for a real, routed, billed request. That is the number a docs-scraping competitor cannot show you, because it only exists once you actually call the endpoint.

The reason this matters for budgeting is that M2's per-token rate is low enough that for short interactive turns, latency and reliability will dominate your experience long before cost does. We re-read the billed-cost field rather than estimating it, and it matched the routed rate card, which is the confirmation we wanted before recommending M2 as a default.

How M2 compares to a cheaper DeepSeek baseline

If you are price-shopping across Chinese models, the honest comparison is against DeepSeek's budget tier, which is cheaper per token than M2.

According to DeepSeek API Docs, DeepSeek V4-Flash is billed at $0.14 input and $0.28 output per million tokens on the official endpoint, well under M2's $0.255 / $1.00 routed rate. Our DeepSeek anchor call (27 input, 86 output) returned in 1.96 seconds against that official endpoint. There is an asymmetry to keep honest here: the M2 figures are OpenRouter-routed while the DeepSeek figures are official-endpoint, so this is a cross-provider directional read, not a like-for-like A/B on one platform.

The takeaway is not that one model wins outright. DeepSeek V4-Flash is the cheaper raw tier, and if your task is well served by it, you will pay less. M2 earns its slightly higher price when you want MiniMax's model behavior specifically, or when MiniMax's higher tiers (M3 for larger inputs) sit in the same account and you want one vendor. For a fuller cross-model rate card and the rest of the MiniMax lineup, see the hub.

When MiniMax-M2 is the right default

Choose M2 if you want a single MiniMax tier that handles both general chat and coding without you having to route per task, and you are comfortable buying it through an aggregator to shave the per-token rate. Buy it on OpenRouter if the roughly 15% input and 17% output discount over the official rate matters at your volume and you do not need native-account features.

Go to the official platform if you need a native MiniMax account, region-specific terms, or anything the aggregator does not expose, and budget at the $0.30 / $1.20 official rate in that case. Look past M2 toward DeepSeek V4-Flash if raw per-token cost is your single deciding factor and the task does not need MiniMax specifically. And step up to M3 inside MiniMax when your inputs are large enough that the higher-context tier earns its price.

FAQ

How much does the MiniMax-M2 API cost? The cheapest verified path is OpenRouter at $0.255 input and $1.00 output per million tokens, measured against our own billed account on 2026-06-26. The official MiniMax rate is $0.30 / $1.20 per million, which is pending a native-key re-verification.

Is MiniMax-M2 cheaper on OpenRouter or on the official platform? On OpenRouter, slightly. The routed rate of $0.255 / $1.00 is about 15% under the official $0.30 / $1.20 on input and 17% under on output. The trade-off is that the official platform gives you a native account and region settings the aggregator route does not.

What did a real MiniMax-M2 call cost? A general prompt of 39 input and 200 output tokens billed $0.000209945 in 5.08 seconds, and a coding prompt of 38 input and 200 output tokens billed $0.0002514 in 2.95 seconds. Both were measured via OpenRouter; the 200-token outputs were capped by our request, not the model.

Is MiniMax-M2 cheaper than DeepSeek? No, not per token. DeepSeek V4-Flash is $0.14 / $0.28 on its official endpoint, below M2's routed $0.255 / $1.00. M2 is worth the premium only when you specifically want MiniMax's model behavior or want to consolidate on one vendor.

Does MiniMax-M2 work for US and international users? Yes. Access is available through the international MiniMax platform, and the OpenRouter route works from the same regions OpenRouter already serves.

This article is part of the MiniMax API pricing hub, which carries the full lineup rate card. For the cheaper Chinese-model baseline referenced above, see the DeepSeek API pricing hub.

Author: Kevin Fan, Customer Success Manager at China LLM Directory, specializing in Chinese LLM ecosystem pricing. Last verified: 2026-06-26.

MiniMax-M2 is MiniMax's default general-and-coding tier, and the cheapest way in is OpenRouter at $0.255 input and $1.00 output per million tokens, a…

Written by Fan Chuanyu

MiniMax-M2 API pricing (verified 2026-06)

What one MiniMax-M2 call actually cost (first-hand, via OpenRouter)

How M2 compares to a cheaper DeepSeek baseline

When MiniMax-M2 is the right default

FAQ