Skip to main content

MiMo-V2.5-Pro (Reasoning)

XiaomiOpen WeightMIT · Commercial OK

Description

MiMo-V2.5-Pro is Xiaomi's 1.02T-parameter sparse Mixture-of-Experts language model with 42B active parameters and a 1M-token context window. It inherits the MiMo-V2-Flash hybrid-attention and Multi-Token Prediction design, extends context during pre-training up to 1M tokens, and uses supervised fine-tuning, domain-specialized reinforcement learning, and Multi-Teacher On-Policy Distillation to improve complex software engineering, long-horizon agentic tasks, and ultra-long-context coherence.

Release Date
2026-04-22
Parameters
1.0T
Context Length
1.0M
Modalities
audio, text

Capability Radar

70
general
70
coding
80
reasoning
60
scienceest.
70
agents
30
multimodal

Science is estimated from LLM Stats science scores or reasoning when no dedicated science benchmarks are available.

Rankings

Domain#RankScoreSource
Agentic Capability69
48.0
LS

Benchmark Scores (LLM Stats)

(LLM Stats (zeroeval))

Agents

TAU3-Bench72.9%SR
WildClawBench43.0%SR
Finance Agent v241.5%

Code

FrontierSWE (Impl.)340.0%SR
MiMo Coding Bench73.7%SR
Claw-Eval64.0%SR

General

C-Eval91.5%SR
MMLU89.4%SR
Global-MMLU83.6%SR
TriviaQA81.3%SR
SWE-bench Verified (Agentless)35.7%SR

Language

MMLU-Redux92.8%SR
CMMLU90.2%SR
MMLU-Pro68.5%SR

Long Context

GraphWalks62.0%SR

Math

GSM8k99.6%SR
MATH86.2%SR
AIMEMAA37.3%SR

Reasoning

ARC-C97.2%SR
HellaSwagAI2 (2019)89.8%SR
BBH88.4%SR
DROP86.3%SR
Winogrande85.6%SR
SWE-Bench Verified78.9%SR
HumanEval+75.6%SR
MBPP+74.1%SR
Terminal-Bench 2.0Stanford × Laude Institute (2026)68.4%SR
GPQANYU + Cohere + Anthropic (2023)66.7%SR
SWE-Bench ProPrinceton NLP (2024)57.2%SR
LiveCodeBench v639.6%SR
Humanity's Last Exam34.0%SR

AA Evaluation Indices

(Artificial Analysis)

No AA evaluation data available

LLM Stats Category Scores

(LLM Stats (zeroeval))
Language
90
Legal
80
Math
80
Reasoning
80
Frontend Development
80
Healthcare
80
Physics
70
Finance
70
General
70
Biology
70
Chemistry
70
Code
70
Tool Calling
70
Long Context
60
Agents
60
Coding
40
Vision
30

Pricing

Input Price$0.348 / 1M tokens
Output Price$0.696 / 1M tokens
Blended Price (3:1)$0.54375 / 1M tokens
Cache Read Price$0.0036 / 1M tokens

Speed

No speed data available

Provider Price Ranking

Provider Price Ranking

3 providers

Cheapest: XiaomiMost Expensive: Novita
ProviderInputOutput
1XiaomiPRIMARY
$0.348
$0.696
2DeepInfra
$1
$3
3Novita
$2
$6

Compare pricing across different API providers for this model.

External Sources