Model reference · open weights

MiMo-Pro

Available as managed deployment LLMs XiaomiMiMo Text gen 1 variants 25k dl/mo

MiMo-Pro is an open-weight language model from XiaomiMiMo. AxForge deploys and operates it for you on dedicated EU-owned hardware — with the licence handled where one is required.

Available as managed deployment — configured and operated for you on dedicated EU hardware, quoted per deployment.

What it is

Released byXiaomiMiMo
TypeLanguage models
TaskText gen
Parameters (lead)1023.2B
Context1024k tokens
Runs withtransformers
Released2026-04-27
Popularity25k downloads / month
LicenceOpen weights

About

What MiMo-Pro is

|  |  |  |  |

 |   |   | 

MiMo-V2.5-Pro

MiMo-V2.5-Pro is an open-source Mixture-of-Experts (MoE) language model with 1.02T total parameters and 42B active parameters. It utilizes the hybrid attention architecture and 3-layers Multi-Token Prediction (MTP) introduced in MiMo-V2-Flash, with up to 1M tokens context length.

1. Introduction

MiMo-V2.5-Pro is our most capable model to date, designed for the most demanding agentic, complex software engineering, and long-horizon tasks. It sustains complex trajectories spanning thousands of tool calls with strong instruction following and coherence over a 1M-token context window. Key features include:

Read the full model card
  • Hybrid Attention Architecture: Interleaves Sliding Window Attention (SWA) and Global Attention (GA) with a 6:1 ratio and 128 sliding window. This reduces KV-cache storage by nearly 7x while maintaining long-context performance via learnable attention sink bias.
  • Multi-Token Prediction (MTP): 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.
  • Efficient Pre-Training: Trained on 27T tokens using FP8 mixed precision and native 32k seq length. The context window supports up to 1M tokens.
  • Agentic Capabilities: Post-training utilizes SFT, large-scale agentic RL and Multi-Teacher On-Policy Distillation (MOPD), achieving superior performance on the most demanding agentic, complex software engineering, and long-horizon tasks.

2. Model Downloads

ModelTotal ParamsActive ParamsContext LengthPrecisionDownload
MiMo-V2.5-Pro1.02T42B1MFP8 (E4M3) Mixed🤗 HuggingFace 🤖 ModelScope
MiMo-V2.5-Pro-Base1.02T42B256KFP8 (E4M3) Mixed🤗 HuggingFace 🤖 ModelScope

3. Evaluation Results

Base Model Evaluation

CategoryBenchmarkSettingMiMo-V2.5-Pro BaseMiMo-V2.5 BaseDeepSeek-V4-Pro BaseDeepSeek-V4-Flash BaseKimi-K2 Base
Params#Activated / #Total-42B / 1.02T15B / 310B49B / 1.6T13B / 284B32B / 1.04T
GeneralBBH3-shot88.487.287.586.988.7
MMLU5-shot89.486.390.188.787.8
MMLU-Redux5-shot92.889.890.889.490.2
MMLU-Pro5-shot68.565.873.568.369.2
DROP3-shot86.383.788.788.683.6
ARC-Challenge25-shot97.296.5--96.2
HellaSwag10-shot89.888.688.085.794.6
WinoGrande5-shot85.684.781.579.585.3
TriviaQA5-shot81.380.785.682.885.1
GPQA-Diamond5-shot66.758.1--48.1
MathGSM8K8-shot99.683.392.690.892.1
MATH4-shot86.267.764.557.470.2
AIME 24&252-shot37.336.9--31.6
CodeHumanEval+1-shot75.671.3--84.8
MBPP+3-shot74.170.9--73.8
LiveCodeBench v61-shot39.635.5--26.3
SWE-Bench (AgentLess)3-shot35.730.8--28.2
ChineseC-Eval5-shot91.588.693.192.192.5
CMMLU5-shot90.288.290.890.490.9
MultilingualGlobalMMLU5-shot83.677.4--80.7

Long-context Evaluation

GraphWalks is a long-context benchmark from OpenAI that fills the prompt with a directed graph of hex-hash nodes and asks the model to run a breadth-first search (nodes exactly at depth N) or list a node's parents. We evaluate across the full 32k–1M input-token span and apply the same evaluation fixes described by Anthropic.

MiMo V2.5 Pro delivers a major leap in long-context reasoning. Past 128k, V2 Pro degrades rapidly and collapses to 0.00 at 1M on both subtasks, while V2.5 Pro still scores 0.56 BFS / 0.92 Parents at 512k and 0.37 / 0.62 at 1M.

4. Model Architecture & Training Process

MiMo-V2.5-Pro addresses the quadratic complexity of long contexts by interleaving Local Sliding Window Attention (SWA) and Global Attention (GA). Unlike traditional speculative decoding, our MTP module is natively integrated for training and inference.

Model Summary

ComponentMiMo-V2.5-ProMiMo-V2.5
Total Parameters1.02T310B
Activated Parameters42B15B
Hidden Size61444096
Num Layers70 (1 dense + 69 MoE)48 (1 dense + 47 MoE)
Full Attention Layers109
SWA Layers6039
Num Attention Heads12864
Num KV Heads8 (GQA)8 (GA) / 4 (SWA)
Head Dim (QK / V)192 / 128192 / 128
Routed Experts384256
Experts per Token88
MoE Intermediate Size20482048
Dense Intermediate Size16384 (layer 0 only)16384 (layer 0 only)
SWA Window Size128128
Max Context Length1M1M
MTP Layers33

Training Process

For post-training, MiMo-V2.5-Pro adopts the three-stage post-training paradigm introduced in MiMo-V2-Flash to achieve exceptional performance. The paradigm begins with Supervised Fine-Tuning (SFT) to build strong, foundational instruction-following skills using curated data pairs. Next, in the Domain-Specialized Training stage, diverse teacher models — ranging from math and safety to complex agentic tool-use — are individually optimized using domain-specific RL rewards. Finally, the process culminates in Mult

From the published model card. Full card on the HuggingFace links in the sidebar.

Using it via the API

Call it like any OpenAI endpoint

Once AxForge deploys mimo-pro for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (mimo-pro below is illustrative; you get the exact model name on deployment.)

$ curl -sS https://api.axforge.ai/v1/chat/completions \
  -H "Authorization: Bearer $AXFORGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"mimo-pro","messages":[{"role":"user","content":"Hello"}]}'

Create an account — your API key is available in the console. 3M free tokens every 30 days with every new account.

© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms