MiMo-V2.5-Pro

Name: MiMo-V2.5-Pro
Author: Xiaomi

Xiaomi

MiMo-V2.5-Pro is Xiaomi's flagship 1T-parameter open-weight MoE model released in April 2026. It is the largest model in the MiMo-V2.5 family and supports a 1M token context window. Weights are available on Hugging Face under the MIT license.

Key Specifications

Parameters

1.0T

Context

1.0M

Release Date

April 27, 2026

Average Score

71.8%

API Documentation Research Paper

Timeline

Key dates in the model's history

Announcement

April 27, 2026

Last Update

August 27, 2026

Today

September 11, 2026

Technical Specifications

Parameters

1.0T

Training Tokens

Knowledge Cutoff

Family

Capabilities

MultimodalZeroEval

Pricing & Availability

Input (per 1M tokens)

$0.43

Output (per 1M tokens)

$0.87

Max Input Tokens

1.0M

Max Output Tokens

131.1K

Supported Features

Function CallingStructured OutputCode ExecutionWeb SearchBatch InferenceFine-tuning

Benchmark Results

Model performance metrics across various tests and benchmarks

General Knowledge

Tests on general knowledge and understanding

HellaSwag

Base model, 10-shot • Self-reported

89.8%

MMLU

Base model, 5-shot • Self-reported

89.4%

Winogrande

Base model, 5-shot • Self-reported

85.6%

Programming

Programming skills tests

SWE-Bench Verified

• Self-reported

78.9%

Mathematics

Mathematical problems and computations

GSM8k

Base model, 8-shot • Self-reported

99.6%

MATH

Base model, 4-shot • Self-reported

86.2%

Reasoning

Logical reasoning and analysis

DROP

Base model, 3-shot • Self-reported

86.3%

GPQA

Base model, 5-shot • Self-reported

66.7%

Other Tests

Specialized benchmarks

AIME

Base model, AIME 24&25, 2-shot • Self-reported

37.3%

ARC-C

Base model, 25-shot • Self-reported

97.2%

BBH

Base model, 3-shot • Self-reported

88.4%

C-Eval

Base model, 5-shot • Self-reported

91.5%

Claw-Eval

Hugging Face evaluation result, General • Self-reported

64.0%

CMMLU

Base model, 5-shot • Self-reported

90.2%

Finance Agent v2

• Verified

41.5%

Global-MMLU

Base model, 5-shot • Self-reported

83.6%

GraphWalks

Parents 1M F1 • Self-reported

62.0%

HumanEval+

Base model, 1-shot • Self-reported

75.6%

Humanity's Last Exam

No tools • Self-reported

34.0%

LiveCodeBench v6

Base model, 1-shot • Self-reported

39.6%

MBPP+

Base model, 3-shot • Self-reported

74.1%

MiMo Coding Bench

• Self-reported

73.7%

MMLU-Pro

Base model, 5-shot • Self-reported

68.5%

MMLU-Redux

Base model, 5-shot • Self-reported

92.8%

SWE-Bench Pro

• Self-reported

57.2%

SWE-bench Verified (Agentless)

Base model, 3-shot • Self-reported

35.7%

TAU3-Bench

• Self-reported

72.9%

Terminal-Bench 2.0

• Self-reported

68.4%

TriviaQA

Base model, 5-shot • Self-reported

81.3%

WildClawBench

Hugging Face evaluation result, Overall • Self-reported

43.0%

License & Metadata

License

mit

Announcement Date

April 27, 2026

Last Updated

August 27, 2026

Similar Models

All Models

MiMo-V2-Flash

Xiaomi

309.0B

Best score:0.8 (GPQA)

Released:Dec 2025

GLM-5

Zhipu AI

744.0B

Best score:0.9 (TAU)

Released:Feb 2026

Kimi K2-Thinking-0905

Moonshot AI

1.0T

Best score:0.8 (GPQA)

Released:Sep 2025

Price:$0.60/1M tokens

GLM-5.1

Zhipu AI

754.0B

Best score:0.9 (GPQA)

Released:Apr 2026

Price:$1.40/1M tokens

GLM-4.7

Zhipu AI

358.0B

Best score:0.9 (TAU)

Released:Dec 2025

Price:$0.60/1M tokens

Kimi K2 Instruct

Moonshot AI

1.0T

Best score:0.9 (HumanEval)

Released:Jan 2025

Price:$0.57/1M tokens

Nemotron 3 Ultra (550B A55B)

NVIDIA

550.0B

Best score:0.9 (GPQA)

Released:Jun 2026

Kimi K2-Instruct-0905

Moonshot AI

1.0T

Best score:0.9 (MMLU)

Released:Sep 2025

Price:$0.60/1M tokens

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.