MiMo-V2.6-Pro

Xiaomi开源权重MIT · 商用许可

描述

MiMo-V2.6-Pro is Xiaomi's MIT-licensed omnimodal sparse Mixture-of-Experts model with 1.02T total parameters and 42B active parameters. It supports text, image, video, and audio understanding with a 1,048,576-token context window and text output. MiMo-V2.6 scales reinforcement learning across coding, general agents, visual tasks, and cybersecurity. Xiaomi's launch article is dated September 22, 2026; released weights are named MiMo-V2.6-Pro-RL. Benchmark results are Xiaomi-reported release evaluations, distinct from intermediate live-training checkpoints. UltraSpeed serves the same model at higher speed for $4.35 input and $8.70 output per million tokens.

发布日期

2026-09-21

参数规模

1.0T

上下文长度

1.0M

支持模态

audio, image, text, video

能力雷达图

general

coding

100

reasoning

science估算

agents

multimodal

缺少专门科学评测时，Science 由 LLM Stats 科学得分或推理能力估算。

排行榜排名

领域	#排名	分数	来源
智能体能力模型榜	74	47.0	LS
代码能力榜	30	82.0	LB
通用能力榜	28	45.0	AA
多模态榜	35	60.0	LS
推理能力	40	80.0	LB
科学能力	32	43.0	AA

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Agents

Toolathlon-Verified

76.9%自报

SEC-bench Pro

66.3%自报

JobBench

62.0%自报

AutomationBench v1.0.6

53.1%自报

Agents' Last Exam

31.6%自报

Program Bench

26.5%自报

ExploitGym

17.8%自报

Code

CyberGym

94.0%自报

MiMo Cyber Bench

81.7%自报

MiMo Visual Coding

72.3%自报

DeepSWE 1.1

71.9%自报

MiMo Coding Bench

63.2%自报

ExploitBench

47.9%自报

General

GDPval-AA 2.1

1673.00 / 3000自报

AA Intelligence Index v4.3

46.3%自报

Multimodal

OSWorld-Verified

82.0%自报

Reasoning

Terminal-Bench 2.1

89.9%自报

Terminal-Bench 4.0

34.9%自报

AA 评测指数

(Artificial Analysis)

Strategy And Ops

46.9

Finance And Accounting

38.9

Economics

36.1

Healthcare And Medical

35.8

Legal

35.0

Engineering(Artificial Analysis)

34.3

Intelligence Index(Artificial Analysis)

29.9

LLM Stats 分类评分

(LLM Stats (zeroeval))

Reasoning

100

Agents

100

General

100

Multimodal

Vision

Knowledge

Productivity

Safety

Code

Tool Calling

定价

输入价格$0.435 / 1M tokens

输出价格$0.87 / 1M tokens

混合价格(3:1)$0.54375 / 1M tokens

缓存读取价格$0.0036 / 1M tokens

速度

Tokens/秒62.3

首Token延迟3.38s

首回答延迟35.49s

供应商价格排行

18 个供应商

最便宜: Requesty最贵: Neuralwatt

供应商输入输出

1Requesty最便宜

$0.43

$0.87

2Standard Compute

$0.43

$0.87

3Xiaomi主要

$0.435

$0.87

4NanoGPT

$0.435

$0.87

5OpenRouter

$0.435

$0.87

6OpenCode Go

$0.435

$0.87

7ZenMux

$0.435

$0.87

8Kilo Gateway

$0.435

$0.87

9Ollama Cloud

$0.435

$0.87

10Vercel AI Gateway

$0.435

$0.87

11DevPass (LLM Gateway)

$0.435

$0.87

12EmpirioLabs AI

$0.435

$0.87

13LLM Gateway

$0.435

$0.87

14Boundless

$0.435

$0.87

15CrossModel

$0.47

$0.94

16AIHubMix

$0.48

$0.96

17above.dev

$0.5077

$1.0154

18Neuralwatt

$0.87

$1.74

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