MiMo-V2.5-Pro (Reasoning)

Xiaomi开源权重MIT · 商用许可

描述

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.

发布日期

2026-04-22

参数规模

1.0T

上下文长度

1.0M

支持模态

audio, text

能力雷达图

general

coding

reasoning

science

agents

multimodal

排行榜排名

领域	#排名	分数	来源
智能体能力模型榜	55	48.0	LS
代码能力榜	137	76.0	AA
通用能力榜	83	69.0	AA
科学能力	104	73.0	AA

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Agents

TAU3-Bench

72.9%自报

WildClawBench

43.0%自报

Finance Agent v2

41.5%

Code

FrontierSWE (Impl.)

340.0%自报

MiMo Coding Bench

73.7%自报

Claw-Eval

64.0%自报

General

C-Eval

91.5%自报

MMLU

89.4%自报

Global-MMLU

83.6%自报

TriviaQA

81.3%自报

SWE-bench Verified (Agentless)

35.7%自报

Language

MMLU-Redux

92.8%自报

CMMLU

90.2%自报

MMLU-Pro

68.5%自报

Long Context

GraphWalks

62.0%自报

Math

GSM8k

99.6%自报

MATH

86.2%自报

AIMEMAA

37.3%自报

Reasoning

ARC-C

97.2%自报

HellaSwagAI2 (2019)

89.8%自报

BBH

88.4%自报

DROP

86.3%自报

Winogrande

85.6%自报

SWE-Bench Verified

78.9%自报

HumanEval+

75.6%自报

MBPP+

74.1%自报

Terminal-Bench 2.0Stanford × Laude Institute (2026)

68.4%自报

GPQANYU + Cohere + Anthropic (2023)

66.7%自报

SWE-Bench ProPrinceton NLP (2024)

57.2%自报

LiveCodeBench v6

39.6%自报

Humanity's Last Exam

34.0%自报

AA 评测指数

(Artificial Analysis)

Tau2(Sierra + U Toronto + Vector Institute (2025))

94.2

Gpqa(NYU + Cohere + Anthropic (2023))

86.6

Ifbench(Google Research (2023))

79.9

Lcr(Artificial Analysis)

79.7

Terminalbench V2 1

65.2

Coding Index(Artificial Analysis)

60.2

Scicode(UIUC + Argonne National Lab (2024))

50.6

Terminalbench Hard(Stanford × Laude Institute (2026))

43.2

Hle(Center for AI Safety + Scale AI (2025))

35.7

Intelligence Index(Artificial Analysis)

26.0

Tau Banking

9.9

Terminalbench V4 0

0.0

LLM Stats 分类评分

(LLM Stats (zeroeval))

Language

Legal

Math

Reasoning

Frontend Development

Healthcare

Physics

Finance

General

Biology

Chemistry

Code

Tool Calling

Long Context

Agents

Coding

Vision

定价

输入价格$0.435 / 1M tokens

输出价格$0.87 / 1M tokens

混合价格(3:1)$0.544 / 1M tokens

缓存读取价格$0.0036 / 1M tokens

速度

Tokens/秒0.0

首Token延迟0.00s

首回答延迟0.00s

供应商价格排行

4 个供应商

最便宜: Xiaomi最贵: EmpirioLabs AI

供应商输入输出

1Xiaomi最便宜

2DeepInfra

3Novita

$0.00001

4EmpirioLabs AI

$2.175

$4.35

比较该模型在不同 API 供应商之间的定价。

外部链接

LLM Stats Artificial Analysis