If you are searching for MiMo V2 Pro, Mimo V2 Pro, Xiaomi MiMo-V2-Pro, or Hunter Alpha, this is the short version:
- the official Xiaomi spelling is MiMo-V2-Pro
- Xiaomi publicly introduced it on March 18, 2026
- it is a flagship agent-focused foundation model, not a video generator
- the official public entry points are AI Studio, Xiaomi's API platform, and OpenRouter
If you actually meant the video product on this site, that is a different product. You can use the Mimo AI Video Generator separately.
If you want the main site entry instead, visit the Mimo AI homepage.
TL;DR
- MiMo-V2-Pro is Xiaomi MiMo's flagship agent model for coding, tool use, search, and long-horizon workflows.
- Xiaomi's official product page says it has more than 1T total parameters, 42B active parameters, and supports up to 1M-token context.
- Xiaomi also says the model ranks 8th worldwide and 2nd among Chinese LLMs on the Artificial Analysis Intelligence Index.
- Hunter Alpha was not a different model family. Xiaomi says it was an early internal test build of MiMo-V2-Pro.
- Xiaomi's public API pricing starts at $1 input / $3 output per million tokens up to 256K context, then $2 / $6 from 256K to 1M.
What Is MiMo V2 Pro?
As of March 20, 2026, the clearest official description comes from Xiaomi's own launch page published on March 18, 2026.
Xiaomi positions MiMo-V2-Pro as its flagship foundation model for real-world agentic workloads. In plain English, that means the model is not being sold as a chat toy or a demo-first assistant. Xiaomi is presenting it as the reasoning engine behind:
- coding agents
- multi-step tool workflows
- long-context task execution
- search and browsing tasks
- production engineering use cases
In practice, most people evaluating MiMo V2 Pro are trying to answer a few concrete questions:
- is it good for coding?
- is it better than MiMo-V2-Flash?
- can I access it through API?
- is Hunter Alpha the same thing?
- is it actually competitive with Claude, Gemini, and GPT?
MiMo V2 Pro Specs Confirmed So Far
These are the product details Xiaomi has publicly confirmed or published through official access pages reviewed on March 20, 2026:
| Item | Confirmed Public Detail |
|---|---|
| Official name | Xiaomi MiMo-V2-Pro |
| Launch date | March 18, 2026 |
| Positioning | flagship foundation model for agentic workloads |
| Total parameters | over 1T |
| Active parameters | 42B |
| Context window | up to 1M tokens |
| Public access | AI Studio, Xiaomi API Platform, OpenRouter |
| OpenRouter max context | 1,048,576 tokens |
| OpenRouter max output | 131,072 tokens |
There is one important caveat here.
Unlike MiMo-V2-Flash, Xiaomi's Pro launch page does not present MiMo-V2-Pro as an open-source release. So if you are specifically searching for a public checkpoint, weights, or Hugging Face download, the public evidence currently points you to web/API access, not a Pro checkpoint download.
MiMo V2 Pro Pricing
Xiaomi's official MiMo-V2-Pro page shows tiered API pricing:
| Range | Input | Output | Cache Read | Cache Write |
|---|---|---|---|---|
| MiMo-V2-Pro up to 256K | $1 | $3 | $0.20 | $0 |
| MiMo-V2-Pro 256K to 1M | $2 | $6 | $0.40 | $0 |
All figures above are per million tokens, and Xiaomi notes that cache write is temporarily free.
That is a meaningful part of the MiMo V2 Pro story because Xiaomi is not only pushing long context. It is trying to make long context relatively affordable compared with premium closed models.
For reference, the same official pricing table on Xiaomi's launch page compares MiMo-V2-Pro against:
- Claude Sonnet 4.6 at $3 input / $15 output
- Claude Opus 4.6 at $5 input / $25 output
OpenRouter's model page also lists $1 per million input tokens and $3 per million output tokens, while separately confirming the 1,048,576-token context window and 131,072-token maximum output.
Hunter Alpha and MiMo V2 Pro
This is one of the highest-value search questions around the model.
According to Xiaomi's official MiMo-V2-Pro page, Hunter Alpha was an early internal test build of MiMo-V2-Pro. Xiaomi says Hunter Alpha appeared on OpenRouter about one week before the formal launch, climbed the daily usage charts, and passed 1T total tokens of usage during that early test period.
So if you are comparing the two names:
- Hunter Alpha = pre-release public test identity
- MiMo-V2-Pro = official product name after launch
That distinction matters because both names still appear in public discussions and model listings.
MiMo V2 Pro Benchmarks and Performance Positioning
Xiaomi is very clearly marketing MiMo-V2-Pro as an agent model first, not just a "reasoning benchmark" model.
On Xiaomi's official launch page, the strongest public benchmark signals are:
| Benchmark | MiMo-V2-Pro | Xiaomi's framing |
|---|---|---|
| Artificial Analysis Intelligence Index | #8 worldwide, #2 among Chinese LLMs | broader model intelligence ranking |
| PinchBench average | 81.0 | #3 globally |
| ClawEval | 61.5 | #3 globally, approaching Claude Opus 4.6 |
| SWE-bench Verified | 78.0 | strong coding-agent performance |
| SWE-bench Multilingual | 71.7 | strong multilingual software engineering |
| Terminal-Bench 2.0 | 57.1 | stronger than MiMo-V2-Flash |
| DeepSearch QA-F1 | 86.7 | strong search-agent performance |
Xiaomi also makes two explicit product claims worth noting:
- coding ability surpasses Claude 4.6 Sonnet
- general agent performance approaches Claude Opus 4.6
Those are Xiaomi's own launch-page claims, not neutral third-party verdicts, so they should be read that way. But they do show exactly how Xiaomi wants the model to be evaluated in the market: not as a cheap general chatbot, but as a serious agent engine.
MiMo V2 Pro vs MiMo V2 Flash
If you are deciding between MiMo V2 Pro and MiMo V2 Flash, the split is relatively clean.
| If you care most about... | Better fit |
|---|---|
| maximum agent capability | MiMo-V2-Pro |
| long-horizon workflows with 1M context | MiMo-V2-Pro |
| lower cost and higher speed | MiMo-V2-Flash |
| open-source access | MiMo-V2-Flash |
| developer experimentation on public checkpoints | MiMo-V2-Flash |
| flagship production agent workloads | MiMo-V2-Pro |
Here is the practical comparison from public sources reviewed on March 20, 2026:
| Category | MiMo-V2-Pro | MiMo-V2-Flash |
|---|---|---|
| Release date | March 18, 2026 | December 16, 2025 |
| Positioning | flagship agent model | fast and efficient open model |
| Parameters | over 1T total, 42B active | 309B total, 15B active |
| Context | 1M | 256K |
| Availability | AI Studio, API, OpenRouter | Hugging Face, AI Studio, API |
| Public status | API/web access | open-source release |
| Entry pricing | $1/$3 up to 256K | $0.1/$0.3 per million |
The simplest decision rule is this:
- choose MiMo-V2-Pro if capability ceiling matters more than cost
- choose MiMo-V2-Flash if speed, price, and public model access matter more
How to Access MiMo V2 Pro
As of March 20, 2026, the public access paths are:
- Xiaomi MiMo AI Studio for direct interactive use
- Xiaomi MiMo API Platform for production integration
- OpenRouter if you want MiMo-V2-Pro through an existing OpenAI-compatible routing layer
That access pattern is another signal that MiMo-V2-Pro is aimed at developers and agent builders. Xiaomi is not hiding it behind a waitlist-only research page. It is already attached to concrete distribution channels.
Who Should Care About MiMo V2 Pro?
MiMo V2 Pro is most relevant if you are building:
- coding agents
- browser or search agents
- long-context workflow assistants
- internal developer tools
- multi-step task automation systems
- tool-using copilots that need better stability
It is less relevant if your main need is:
- the cheapest general-purpose chatbot
- an open checkpoint you can self-host immediately
- image or video generation
- lightweight consumer chat use
FAQ
Is MiMo V2 Pro open source?
Not based on the public MiMo-V2-Pro launch page reviewed on March 20, 2026. Xiaomi provides web and API access, but the page does not present Pro as an open-source release. MiMo-V2-Flash is the public open-source release in the current MiMo V2 lineup.
Is Hunter Alpha the same as MiMo V2 Pro?
Yes. Xiaomi says Hunter Alpha was an early internal test build of MiMo-V2-Pro that appeared on OpenRouter before the official launch.
How much does MiMo V2 Pro cost?
According to Xiaomi's official page, pricing starts at $1 input / $3 output per million tokens up to 256K context, then $2 / $6 for 256K to 1M context.
What is MiMo V2 Pro's context window?
Xiaomi's launch page says the model supports up to 1M-token context. OpenRouter's model page lists that as 1,048,576 tokens with a 131,072-token maximum output.
Is MiMo V2 Pro better than MiMo V2 Flash?
For raw flagship agent capability, yes, that is how Xiaomi positions it. For cost, speed, and open-source access, MiMo-V2-Flash is usually the better fit.
Is MiMo V2 Pro related to Mimo AI video generator?
No. Xiaomi MiMo-V2-Pro is a language model for agentic workloads. Mimo AI Video Generator on this site is a separate video product.
Sources
Sources reviewed on March 20, 2026: