跳转到主要内容
AITier

MiMo-V2-Flash (Non-reasoning)

Xiaomi开源权重MIT · 商用许可

描述

MiMo-V2-Flash is a powerful, efficient, and ultra-fast foundation language model that excels in reasoning, coding, and agentic scenarios. It is a Mixture-of-Experts model with 309B total parameters and 15B active parameters, featuring a hybrid attention architecture with sliding-window and full attention (5:1 ratio, 128-token window). Delivers 150 tokens/sec inference with 256k context window.

发布日期
2025-12-16
参数规模
309.0B
上下文长度
262K
支持模态
text

能力雷达图

32
general
46
coding
67
reasoning
48
science
60
agents
0
multimodal

排行榜排名

领域#排名分数来源
代码能力榜303
49.0
AA
通用能力榜225
51.0
AA
科学能力379
39.0
AA

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

General

Arena-Hard v286.2%自报
Tau-bench80.3%自报

Language

MMLU-Pro84.9%自报

Long Context

LongBench v260.6%自报
MRCR45.7%自报

Math

AIME 202594.1%自报
HMMT 202584.4%自报

Reasoning

GPQANYU + Cohere + Anthropic (2023)83.7%自报
LiveCodeBench v680.6%自报
SWE-Bench Verified73.4%自报
SWE-bench Multilingual71.7%自报
BrowseCompOpenAI (2025)58.3%自报
Terminal-Bench 2.0Stanford × Laude Institute (2026)38.5%自报
Terminal-Bench30.5%自报
Humanity's Last Exam22.1%自报

AA 评测指数

(Artificial Analysis)
Tau2(Sierra + U Toronto + Vector Institute (2025))
83.9
Mmlu Pro(TIGER-Lab (Univ. of Waterloo, Toronto, CMU, 2024))
74.4
Math Index(Artificial Analysis)
67.7
Aime 25(MAA (Mathematical Association of America))
67.7
Gpqa(NYU + Cohere + Anthropic (2023))
65.6
Terminalbench V2 1
61.8
Coding Index(Artificial Analysis)
49.8
Livecodebench(UC Berkeley + MIT + Cornell (2024))
40.2
Ifbench(Google Research (2023))
39.9
Lcr(Artificial Analysis)
36.3
Terminalbench Hard(Stanford × Laude Institute (2026))
25.8
Intelligence Index(Artificial Analysis)
16.0
Hle(Center for AI Safety + Scale AI (2025))
8.6

LLM Stats 分类评分

(LLM Stats (zeroeval))
Creativity
90
Writing
90
Language
80
Legal
80
Physics
80
Finance
80
Healthcare
80
Biology
80
Chemistry
80
Math
70
Reasoning
70
Frontend Development
70
General
70
Search
60
Structured Output
60
Tool Calling
60
Long Context
50
Agents
50
Code
50
Vision
20

定价

输入价格免费
输出价格免费
混合价格(3:1)免费
缓存读取价格$0.0028 / 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

供应商价格排行

2 个供应商

最便宜: Qiniu最贵: Xiaomi
供应商输入输出
1Qiniu最便宜
$0.1
$0.3
2Xiaomi
$0.14
$0.28

比较该模型在不同 API 供应商之间的定价。

外部链接

LLM StatsArtificial Analysis