Xiaomi MiMo (샤오미 미모)

Name: Xiaomi MiMo (샤오미 미모)
Brand: Xiaomi (샤오미)

Xiaomi (샤오미), China, mimo.xiaomi.com

A 1 trillion-parameter-class multimodal open model that Xiaomi released under the MIT license. The V2.6 series came out in September 2026, and the API is also very cheap

Open Models Developer API

Go to official site

Request a correction

CurrencyUSD KRW EUR

Personal starting priceCheck siteNo paid plan

Free tierFree plan available

Korean supportLimited

Information updated 2026-10-22, source: mimo.xiaomi.com. Prices and benefits may change. Check the official site before you pay.

Pros

Free commercial use under the MIT license
Native multimodal that takes text, images, video and audio all as input
1 million token context
Artificial Analysis intelligence index of 46.32, the highest among open models (official announcement)
Very cheap API prices

Cons

Pro has 1 trillion parameters, so serving it yourself needs a large GPU cluster
No official Korean performance data
Token Plan prices could not be confirmed on public pages
It is a Chinese company's service, so the data policy needs checking

These are opinions compiled by Pickx from public sources.

At a glance

Xiaomi MiMo is Xiaomi's AI model brand. The MiMo-V2.6 series, released on September 22, 2026, is a multimodal model that handles text, images, video and audio in one model, and its weights were released on Hugging Face under the MIT license. Xiaomi announced that the Pro model scored the highest among open models on the Artificial Analysis intelligence index, and its very low API prices became a talking point. You can also use it through the web (MiMo Studio), a coding tool (MiMo Code) and a desktop app.

Full model list

Model	Type	Features	Plans that can use it
MiMo-V2.6-Pro	Flagship multimodal model (MoE)	1.02T total 42B active 1 million token context text/image/video/audio input	Free weights API pay-as-you-go Token Plan
MiMo-V2.6-Flash	Efficient multimodal model	About 311B balanced for performance, speed and cost	Free weights API pay-as-you-go Token Plan
MiMo-V2.6-Pro-UltraSpeed	High-speed serving mode	Up to 20x faster output at the same quality as Pro (official announcement)	API pay-as-you-go
MiMo-V2.6-Distill-Qwen-9B	Small distilled model	9B image+text input small enough to run on a personal PC	Free weights

Plans

Plan	Price	What you can do	Usage limit	Features
Open weights	Free (MIT)	Download from Hugging Face and run on your own servers	Official limits not disclosed	Commercial use allowed
API pay-as-you-go	Flash $0.14/$0.28 Pro $0.435/$0.87 UltraSpeed $4.35/$8.7 (input/output per 1 million tokens)	MiMo API platform (platform.xiaomimimo.com) / batch API	Official limits not disclosed	Large input price discount on cache hits same price as V2.5
Token Plan Team Token Plan	Unverified	Heavy use at predictable cost	Official limits not disclosed	Price could not be confirmed on the official page

Who should choose this plan

Open weights: Companies and research teams that want to run a multimodal model covering images, video and audio on their own servers. The 9B distilled model for a small PC.
API pay-as-you-go (Flash): Developers who want to save as much cost as possible on large-volume processing.
API pay-as-you-go (Pro): Coding and agent work that needs top-tier open model performance at a low price.
General users who care about Korean conversation quality should test it themselves before using it.

Notes

Information checked on: 2026-10-13
The performance figures (intelligence index 46.32, DeepSWE v1.1 Pro 72.57) are Xiaomi's own announcements.
Sources: MiMo-V2.6 official announcement, Hugging Face MiMo-V2.6-Pro model card, Hugging Face XiaomiMiMo