---
source: 'https://howaiworks.ai/models/mimo'
section: models
title: MiMo-V2.5-Pro
description: >-
  Xiaomi's open-weight flagship, released April 22, 2026. A
  1.02-trillion-parameter MoE with 42B active, a 1M-token context, and an MIT
  license.
tags:
  - MiMo
  - Xiaomi
  - Language Model
  - Large Language Model
  - MoE
  - Open Weights
  - MIT License
  - Agentic AI
  - Long Context
  - Chinese AI
  - Latest
category: Language Models
developer: Xiaomi
developerWebsite: 'https://mimo.xiaomi.com/'
modelType: Language Model
releaseDate: '2026-04-22'
lastUpdated: '2026-07-08'
license: MIT
contextWindow: 1M tokens
inputPrice: $0.435
outputPrice: $0.87
officialWebsite: 'https://mimo.xiaomi.com/mimo-v2-5-pro'
docsPage: 'https://huggingface.co/XiaomiMiMo/MiMo-V2.5-Pro'
pricingPage: 'https://mimo.mi.com/docs/en-US/api/pricing'
---

# MiMo-V2.5-Pro

> Xiaomi's open-weight flagship, released April 22, 2026. A 1.02-trillion-parameter MoE with 42B active, a 1M-token context, and an MIT license.

## Overview

MiMo-V2.5-Pro is Xiaomi's open-weight flagship [large language model](https://howaiworks.ai/glossary/large-language-model), **released on April 22, 2026**, with open weights published to Hugging Face five days later on April 27, 2026. Xiaomi's model card describes it as "an open-source Mixture-of-Experts (MoE) language model with 1.02T total parameters and 42B active parameters," using "the hybrid attention architecture and 3-layers Multi-Token Prediction (MTP) introduced in MiMo-V2-Flash, with up to 1M tokens context length."

The headline is the license. A consumer-electronics manufacturer has published a trillion-parameter model under **MIT** — an OSI-approved license with no field-of-use clause, no geographic carve-out, and no acceptable-use addendum. Xiaomi's release note frames it as permitting "free commercial use, secondary training, and fine-tuning without additional authorization." The Hugging Face API returns `license:mit`, and arena.ai's leaderboard metadata independently records the license as `MIT`. It is not the largest permissively licensed model, though — Artificial Analysis's dataset records DeepSeek V4 Pro (1.6T) and Ling-2.6-1T (1.03T) as MIT as well.

The pitch is agentic endurance at low token cost. Xiaomi's launch post claims that "when paired with a proper harness, V2.5-Pro can sustain complex, long-horizon tasks spanning more than a thousand tool calls," and its product page asserts "Peak Agent Performance: Rivals Claude Opus 4.6 in demanding agentic workloads." Xiaomi's release note further claims the model "ranks first among open-source models globally on the GDPVal-AA and ClawEval leaderboards." These are vendor claims; the independent numbers below are more mixed.

### A note on naming

**MiMo is no longer a 7B model.** Xiaomi's first MiMo release, [`XiaomiMiMo/MiMo-7B-RL`](https://huggingface.co/XiaomiMiMo/MiMo-7B-RL), landed on Hugging Face on April 29, 2025 with 7,833,409,536 parameters, and was positioned as a small reasoning model. Much of the writing still indexed under "Xiaomi MiMo" describes that model.

The current flagship shares only the brand. MiMo-V2.5-Pro reports **1,023,244,718,976 total parameters** in its Hugging Face safetensors index — about 131× the 7B model. Two further sources of confusion:

- **MiMo-V2.5** (no "-Pro") is a *different, smaller, omnimodal* model: 310B total / 15B active, with vision and audio encoders. It is not a variant of the Pro.
- **The MiMo-V2 series was deprecated on June 30, 2026.** Xiaomi's site states the V2 models "have been deprecated on June 30. Please migrate to the V2.5 series soon." Benchmarks or token-share statistics attributed to "MiMo-V2-Pro" describe a retired model.

## Capabilities

- **Long-horizon agentic execution**: Xiaomi reports the model building a complete SysY compiler in Rust — "finished in 4.3 hours across 672 tool calls, scoring a perfect 233/233 against the course's hidden test suite" — and an 8,192-line desktop video editor "over 1,868 tool calls across 11.5 hours of autonomous work."
- **Token efficiency**: "On ClawEval, V2.5-Pro lands at 64% Pass^3 using only ~70K tokens per trajectory — roughly 40–60% fewer tokens than Claude Opus 4.6, Gemini 3.1 Pro, and GPT-5.4 at comparable capability levels."
- **Hybrid attention for cheap long context**: The model card states the SWA/GA interleave "reduces KV-cache storage by nearly 7x while maintaining long-context performance via learnable attention sink bias."
- **Multi-Token Prediction**: "Equipped with three lightweight MTP modules using dense FFNs. This triples output speed during inference and will be good to accelerate rollout in RL training."
- **Genuine 1M-token retrieval**: On OpenAI's GraphWalks benchmark, Xiaomi reports that where its predecessor "collapses to 0.00 at 1M on both subtasks," V2.5-Pro "still scores 0.56 BFS / 0.92 Parents at 512k and 0.37 / 0.62 at 1M."
- **Reasoning by default**: The chat template enables `<think>` blocks unless `enable_thinking` is set false. Artificial Analysis classifies it as a reasoning model.
- **Native tool-call format**: A custom XML `<tool_call>` grammar, with dedicated `mimo` reasoning and tool-call parsers in SGLang.

## Technical Specifications

- **Hugging Face repository**: `XiaomiMiMo/MiMo-V2.5-Pro` (base model: `XiaomiMiMo/MiMo-V2.5-Pro-Base`)
- **Total parameters**: 1,023,244,718,976 (1.02T), read from the safetensors index
- **Active parameters**: 42B per token
- **Layers**: 70 — 1 dense + 69 MoE
- **Experts**: 384 routed, top-8 activated, no shared experts
- **Attention**: [Hybrid](https://howaiworks.ai/glossary/attention-mechanism) — 60 Sliding Window Attention layers + 10 Global Attention layers (6:1), 128-token window; 128 query heads, 8 KV heads (GQA); head dim 192 (QK) / 128 (V)
- **MTP layers**: 3
- **Architecture**: [Mixture-of-Experts](https://howaiworks.ai/glossary/mixture-of-experts) [transformer](https://howaiworks.ai/glossary/transformer), `model_type: mimo_v2`, custom modelling code (`trust_remote_code`)
- **Precision**: FP8 (E4M3) mixed — roughly 1.03 TB of [weights](https://howaiworks.ai/glossary/weights) on disk
- **Context window**: 1M tokens (`max_position_embeddings: 1048576`); the Base checkpoint is 256K
- **Vocabulary**: 152,576
- **Pre-training**: "Trained on 27T tokens using FP8 mixed precision and native 32k seq length"
- **Post-training**: SFT, large-scale agentic [reinforcement learning](https://howaiworks.ai/glossary/reinforcement-learning), and Multi-Teacher On-Policy Distillation (MOPD)
- **License**: MIT

Xiaomi does not publish a knowledge cutoff for MiMo-V2.5-Pro — Artificial Analysis records the field as null — so this page does not state one.

## Use Cases

- **Autonomous coding agents**: The model is tuned for harnesses like Claude Code, OpenCode, and Kilo; Xiaomi's own examples run for hours across hundreds of [tool calls](https://howaiworks.ai/glossary/function-calling).
- **High-token agentic workloads**: At $0.435 in / $0.87 out, trajectories that burn tens of millions of tokens become affordable. Pareekh Jain, quoted by InfoWorld, frames the model as a "cost-efficient agent model for high-token workloads."
- **Whole-repository and long-document reasoning**: A 1M-token [context window](https://howaiworks.ai/glossary/context-window) with measured 1M-token retrieval, rather than a nominal one.
- **Self-hosted commercial deployment**: MIT permits redistribution and commercial [inference](https://howaiworks.ai/glossary/inference) without a separate agreement — relevant where a restricted community license is a legal blocker.
- **Distillation and continued pre-training**: The Base checkpoint is published under the same license, making [knowledge distillation](https://howaiworks.ai/glossary/knowledge-distillation) and domain [fine-tuning](https://howaiworks.ai/glossary/fine-tuning) legally straightforward.
- **Latency-tolerant batch work**: Artificial Analysis measures a median time to first answer token of 48.2 seconds — the model thinks at length before answering.

## Performance / Benchmarks

### Independently measured (Artificial Analysis)

Artificial Analysis runs its own evaluations. Its Intelligence Index v4.1 combines nine benchmarks: "GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR."

| Metric | MiMo-V2.5-Pro |
|---|---|
| **Intelligence Index (v4.1)** | **42.2** |
| Coding Index | 60.2 |
| Agentic Index | 29.1 |
| GPQA Diamond | 86.6% |
| IFBench | 79.9% |
| AA-LCR (long-context reasoning) | 73.3% |
| Terminal-Bench v2.1 | 65.2% |
| SciCode | 50.2% |
| Terminal-Bench Hard | 43.2% |
| Humanity's Last Exam | 33.8% |
| 𝜏³-Banking | 8.7% |
| CritPt | 4.0% |
| Apex-Agents | 2.4% |
| AA-Omniscience | 3.6 (hallucination rate 24.5%) |

For scale, Artificial Analysis's own comparison chart places MiniMax-M3 at 44.4, DeepSeek V4 Pro at 44.3, [Kimi K2.6](https://howaiworks.ai/models/kimi) at 44.2, MiMo-V2.5-Pro at 42.2, and [GLM-5.2](https://howaiworks.ai/models/glm) at 51.1. MiMo-V2.5-Pro is competitive with, not ahead of, the open-weight field — while costing considerably less per token.

### Vendor-reported

Xiaomi's post-training benchmark comparison is published **only as an image** (`assets/benchmark.jpg`) on the model card, with no machine-readable table, so this page does not transcribe it. The vendor's stated results are:

- `mimo-v2.5-pro` "ranks first among open-source models globally on the GDPVal-AA and ClawEval leaderboards."
- ClawEval: "64% Pass^3 using only ~70K tokens per trajectory."
- GraphWalks at 1M tokens: 0.37 BFS / 0.62 Parents.

The base-model table on the model card (5-shot MMLU 89.4, 5-shot GPQA-Diamond 66.7, 8-shot GSM8K 99.6, 1-shot LiveCodeBench v6 39.6) describes `MiMo-V2.5-Pro-Base`, not the instruct model, and Xiaomi selected the comparison set. Treat all of the above as vendor claims.

### Human preference (arena.ai)

arena.ai — the former LMArena — ranks `mimo-v2.5-pro` **#29 on its text leaderboard with an Elo of 1466.2** across 34,468 votes (rank interval #15–#39). It carries no rank on arena.ai's webdev leaderboard. Artificial Analysis and arena.ai use unrelated scales and must not be cross-compared.

## Limitations

- **Hard agentic benchmarks stay near the floor.** Despite the long-horizon marketing, Artificial Analysis measures 8.7% on 𝜏³-Banking, 4.0% on CritPt, and 2.4% on Apex-Agents.
- **A 24.5% measured hallucination rate.** AA-Omniscience scores the model at 3.6 with accuracy of 22.6%. Ground factual queries with retrieval.
- **The default system prompt enforces PRC law.** The chat template baked into `tokenizer_config.json` injects a system prompt whenever the caller supplies none, and it states: "You are a **Chinese AI model** and must strictly comply with all applicable laws and regulations of the **People’s Republic of China**. Do not generate, assist with, or facilitate any content that violates Chinese law." It also asserts a fixed identity ("You have **1T parameters**"). Passing an explicit system message overrides it.
- **Vendor benchmarks are unauditable.** The post-training comparison exists only as a JPEG. There is no published table, no per-benchmark methodology, and no independent replication of the "first among open-source models" claim.
- **Serving cost is real.** ~1.03 TB of FP8 weights. Xiaomi's own SGLang example runs `--tp-size 16 --ep-size 16` across multiple nodes. MIT does not make this cheap to host.
- **1M context is not universal.** Among [OpenRouter](https://howaiworks.ai/ai-tools/openrouter) providers, DigitalOcean caps the window at 87,040 tokens and DeepInfra charges $1.00 / $3.00 — more than double Xiaomi's first-party rate.
- **Open weights, closed data.** MIT covers the weights and code. Training data, data mixture, and the RL environments are not released.
- **Enterprise headwinds.** Omdia's Lian Jye Su, quoted by InfoWorld, notes that "closed frontier models may still win on generic tasks, and the hardest edge cases," and that "Chinese-origin models can trigger concerns in regulated Western organizations."

## Pricing & Access

**Self-hosting is free** under MIT: [`XiaomiMiMo/MiMo-V2.5-Pro`](https://huggingface.co/XiaomiMiMo/MiMo-V2.5-Pro) on Hugging Face, mirrored on [ModelScope](https://modelscope.cn/models/XiaomiMiMo/MiMo-V2.5-Pro).

Xiaomi publishes the same rate card in **two currencies**, one per locale. The USD figures are a fixed conversion of the CNY ones at 0.145 USD per yuan — ¥3 × 0.145 = $0.435, ¥6 × 0.145 = $0.87 — so they are not independently set prices.

| Tier | English page (USD / MTok) | Chinese page (CNY / MTok) |
|---|---|---|
| 输入（缓存命中） Input, cache hit | $0.0036 | ¥0.025 |
| 输入（缓存未命中） Input, cache miss | **$0.435** | **¥3** |
| 输出 Output | **$0.87** | **¥6** |

A premium **MiMo-V2.5-Pro-UltraSpeed** tier — "Combines FP4 lossless quantization and DFlash parallel decoding," advertised at "1,000 tokens/s peak" — is priced at $0.0108 / $1.305 / $2.61 (¥0.075 / ¥9 / ¥18), and is early-access only. Its FP4 weights are published as `XiaomiMiMo/MiMo-V2.5-Pro-FP4-DFlash`.

**Third-party hosting** via OpenRouter as `xiaomi/mimo-v2.5-pro`:

| Provider | Input / Output (per 1M) | Context |
|---|---|---|
| Xiaomi (first-party) | $0.435 / $0.87 | 1,048,576 |
| AtlasCloud | $0.435 / $0.87 | 1,024,000 |
| Novita | $0.522 / $1.044 | 1,048,576 |
| DigitalOcean | $0.60 / $3.00 | 87,040 |
| DeepInfra | $1.00 / $3.00 | 1,048,576 |

**Consumer access** runs through [Xiaomi MiMo Studio](https://aistudio.xiaomimimo.com) and a "MiMo Claw" subscription, advertised as a limited-time ¥14.9/month offer bundling native OpenClaw integration.

**Serving:** officially supported by both [SGLang](https://docs.sglang.io/cookbook/autoregressive/Xiaomi/MiMo-V2.5) and [vLLM](https://recipes.vllm.ai/XiaomiMiMo/MiMo-V2.5-Pro). Xiaomi recommends `temperature=1.0`, `top_p=0.95`.

## Ecosystem & Tools

- **[XiaomiMiMo/MiMo-V2.5-Pro on Hugging Face](https://huggingface.co/XiaomiMiMo/MiMo-V2.5-Pro)** — MIT weights, architecture table, deployment guides. 101,692 downloads in the 30 days to July 8, 2026, and 698 cumulative likes.
- **[MiMo-V2.5-Pro launch post](https://mimo.xiaomi.com/mimo-v2-5-pro)** — the SysY compiler, video editor, and analog-EDA case studies
- **[Xiaomi MiMo API Platform](https://platform.xiaomimimo.com/)** — first-party API and Token Plan
- **[vLLM recipe](https://recipes.vllm.ai/XiaomiMiMo/MiMo-V2.5-Pro)** and **[SGLang cookbook](https://docs.sglang.io/cookbook/autoregressive/Xiaomi/MiMo-V2.5)** — the two vendor-endorsed serving paths
- **[OpenRouter `xiaomi/mimo-v2.5-pro`](https://openrouter.ai/xiaomi/mimo-v2.5-pro)** — five providers, first-party included
- **`XiaomiMiMo/MiMo-V2.5`** — the 310B/15B omnimodal sibling (208,300 downloads/30d); **`XiaomiMiMo/MiMo-V2.5-ASR`** is published alongside it, while the MiMo-V2.5-TTS series ships only as an API product — it has no Hugging Face repository
- **Harness support** — Claude Code, OpenCode, and Kilo, per Xiaomi's launch post

## Community & Resources

- [MiMo-V2.5-Pro](https://mimo.xiaomi.com/mimo-v2-5-pro) — Xiaomi's launch post, dated April 27, 2026
- [Xiaomi MiMo-V2.5 series open-sourced](https://mimo.mi.com/docs/en-US/news/latest/v2.5-open-sourced) — the MIT License announcement
- [MiMo-V2.5-Pro on Artificial Analysis](https://artificialanalysis.ai/models/mimo-v2-5-pro) — independent Intelligence Index of 42.2 (v4.1)
- [Xiaomi releases MIT-licensed MiMo models for long-running AI agents](https://www.infoworld.com/article/4164184/xiaomi-releases-mit%E2%80%91licensed-mimo-models-for-long%E2%80%91running-ai-agents.html) — InfoWorld, with analyst caveats
- [Xiaomi open-sources MiMo-V2.5 Pro under MIT license](https://www.thestandard.com.hk/china/article/330592/Xiaomi-open-sources-MiMo-V25-Pro-model-under-MIT-license-topping-benchmarks) — The Standard
- [arena.ai text leaderboard](https://arena.ai/leaderboard/text) — #29, Elo 1466.2
- Compare with [DeepSeek V4](https://howaiworks.ai/models/deepseek), [Kimi K2.6](https://howaiworks.ai/models/kimi), [GLM-5.2](https://howaiworks.ai/models/glm), [Ling-2.6-1T](https://howaiworks.ai/models/ling), [MiniMax-M3](https://howaiworks.ai/models/minimax-m3), and [LongCat-2.0](https://howaiworks.ai/models/longcat)

## Frequently Asked Questions

### Isn't MiMo a 7-billion-parameter model?

It was. Xiaomi's first MiMo release, `XiaomiMiMo/MiMo-7B-RL`, was published to Hugging Face on April 29, 2025 and carries 7,833,409,536 parameters. The current flagship, MiMo-V2.5-Pro, is a Mixture-of-Experts model with 1,023,244,718,976 total parameters — roughly 131 times larger. Coverage that describes MiMo as a small reasoning model is describing the 2025 lineage, not the current flagship.

### When was MiMo-V2.5-Pro released?

April 22, 2026 for API access — the date recorded in Artificial Analysis's model dataset and the creation timestamp of the `xiaomi/mimo-v2.5-pro` listing on OpenRouter. Open weights followed on April 27, 2026, when the Hugging Face repository was created, alongside Xiaomi's blog post dated "April 27th, 2026" stating "Today, we are releasing and open-sourcing MiMo-V2.5-Pro."

### Is MiMo-V2.5-Pro really MIT licensed?

Yes. The Hugging Face API returns `license:mit` for `XiaomiMiMo/MiMo-V2.5-Pro`, and Xiaomi's own release note describes the MIT License as permitting "free commercial use, secondary training, and fine-tuning without additional authorization." MIT is an OSI-approved license with no field-of-use or geographic restrictions. It is not, however, the largest MIT-licensed model: Artificial Analysis's dataset records DeepSeek V4 Pro at 1.6T and Ling-2.6-1T at 1.03T, both MIT. The weights are open; the training data is not published.

### How much does MiMo-V2.5-Pro cost?

Xiaomi's own pricing page lists $0.435 per million input tokens (cache miss), $0.87 per million output tokens, and $0.0036 per million cached input tokens. The same page shows a separate CNY rate card of ¥3 / ¥6 / ¥0.025 on its Chinese locale. Self-hosting is free under MIT.

### Is its Artificial Analysis Intelligence Index score 54?

No — not on the current index. Artificial Analysis's published dataset gives MiMo-V2.5-Pro an Intelligence Index of 42.2 (42.2394, v4.1, marked as measured rather than estimated). The "54" comes from Artificial Analysis's own article of June 8, 2026, which scored MiniMax-M3 at 55, "just ahead of open weights peers Kimi K2.6 (54) and MiMo-V2.5-Pro (54)." That article names no index version and predates v4.1, and scores across Artificial Analysis index versions are not comparable. On the current v4.1 chart the value near 54 belongs to GPT-5.5, at 54.8.

### What is the architecture?

A 70-layer Mixture-of-Experts transformer (1 dense layer plus 69 MoE layers) with 384 routed experts, 8 activated per token, and no shared experts. It interleaves Sliding Window Attention and Global Attention at a 6:1 ratio with a 128-token window, and carries three Multi-Token Prediction layers. Weights ship in FP8 (E4M3) mixed precision.

### How long is the context window?

1M tokens. The `config.json` sets `max_position_embeddings` to 1048576, and Xiaomi's model card reports GraphWalks long-context results across "the full 32k–1M input-token span." Note that not every hosted provider serves the full window — DigitalOcean's OpenRouter endpoint caps at 87,040 tokens.

### Is Xiaomi the largest model provider on OpenRouter?

This cannot be confirmed. Several secondary sources report that Xiaomi's models held roughly 21% of OpenRouter traffic, but `openrouter.ai/rankings` is rendered client-side and its HTML contains no model or provider names at all — not even "anthropic" or "deepseek". Treat any token-share percentage as an unverifiable third-party estimate. The figures in circulation are also attached to MiMo-V2-Pro, the previous flagship, which Xiaomi retired on June 30, 2026.

### What are the model's weakest results?

Artificial Analysis measures 8.7% on 𝜏³-Banking, 4.0% on CritPt, and 2.4% on Apex-Agents. Its AA-Omniscience Index is 3.6, with a measured hallucination rate of 24.5%. Long-horizon agentic competence is the model's headline claim, but the hardest independent agentic and physics benchmarks remain near the floor.

## Related

### Related models

- [DeepSeek V4](https://howaiworks.ai/models/deepseek)
- [Kimi K3](https://howaiworks.ai/models/kimi)
- [GLM-5.3](https://howaiworks.ai/models/glm)
- [Ling-2.6-1T](https://howaiworks.ai/models/ling)
- [MiniMax-M3](https://howaiworks.ai/models/minimax-m3)
- [LongCat-2.0](https://howaiworks.ai/models/longcat)

### Related tools

- [OpenRouter](https://howaiworks.ai/ai-tools/openrouter)

---

Source: https://howaiworks.ai/models/mimo — HowAIWorks.ai