cxuan-ai-labs

Models, Research & Prompt ·

Mi has to support it!

English translation of the Chinese original. This version is generated for international readers and may be refined over time.

Interactive reading view ↗

English | Chinese Original

English translation of the Chinese original. This version is generated for international readers and may be refined over time.

Date: 2026-04-28

This morning, Mi Mi announced MiMo-V2. 5, published URL: https://mimo.xiaomi.com/mimo-v2-5-pro/

image-20260428152153311

And then, after that, a message was sent by Roosevelt of Mimo's big model saying that Mi had opened up Mimo's Token scale and that it went straight to 100T, and the countdown was a month. (But I felt that the sooner people applied, the better, the less they were.)

image-20260428151450799

To be honest, the first reaction is that it's a little ahead of expectations.

Application Linkhttps://100t.xiaomimimo.com/.

When you see this url, you don't have to look at the crap behind me. Just close your eyes!

image-20260428151546198

This is a benefit specifically for AI application developers.

In other words, you can apply for this token amount through the Github project, etc., as long as you have developed it.

image-20260428151649871

I filled out the application in the morning and got a message from Mi by noon.

image-20260428151943535

Then connect the mailbox to the official Mi platform. https://account.xiaomi.com/fe/service/account?_locale=zh_CN.

I was connected, and I was lucky to get the Pro package in the subscription management, about 700 million Credits. This Credits does not directly equal 1: 1 Token, and it has its own set of conversion rules.

image-20260428151341384

I looked at it, and according to the rice Credits package, Pro was equivalent to 329, after Max.

image-20260428152436127

I'm in the official group, and I've seen a small number of places, but someone has got a Max package, which I guess is defined by the amount of consumption that you upload.

But the wave has to be praised for the wave.

It's not the first time that millo-V2. 0 started the free gift to Credits, but before that, Mimo-V2. 0 was the first wave, and because of the enthusiasm, I remember it was delayed.


My judgment

I guess there's a couple of reasons why.

** Now the big model API market is rolling, OpenAI, Google, Anthropic has a bunch of competitors in the country. New players come up with the most direct leverage.

** Two is ecological binding. ** Sufficiently, users will use it; when used, the tool chain, documentation, community follow. Token scale is the entrance, not the end.

** Three is that modelling capacity has come up. ** MiMo-V2. 5 has been on the top of the global open source model before. Open Token is essentially selling model capabilities, not the concept of selling.

So it's not just marketing, it's really something to sell.

It's been accessed.

The response rate was found to be much more stable than the free slot. The peak period does not drop and requests are made for basic seconds. Token consumption is statistically clear, and call records can be exported and used to be more economical.

There was no sudden tightening. Some platforms had been used before and the amount of money had begun to shrink quietly, and that was now clear.

This wave must be great.

The free slot is good for taste. Try the model.

The Pro file is suitable for users with ongoing needs. Token is big enough, dispatch priority, and stable. If you're doing AI application development, product prototype validation, or running automated workflows, the value of this slot is reasonable compared to the current one.

Of course, the final value is not worth your own use. But at least this time, the choice given by millet is serious, not the operation of "open but limited, limited but price increases."

At least I can use it.


The style of the article is based on Layweout.

Layweout is a public-name layout tool I'm making myself, and it's my first product on line.

You can have direct access. https://layweout.vercel.app/For free.