Mistral Small 3.2
MistralMistral开源权重Apache 2.0 · 商用许可
描述
Mistral-Small-3.2-24B-Instruct-2506 is a minor update of Mistral-Small-3.1-24B-Instruct-2503.
发布日期
2025-06-20
参数规模
23.6B
上下文长度
128K
支持模态
image, text
能力雷达图
26
general
19
coding
40
reasoning
34
science
34
agents
90
multimodal
排行榜排名
基准测试分数 (LLM Stats)
(LLM Stats (zeroeval))Factuality
SimpleQA
12.1%自报
General
IF
84.8%自报
MMLU
80.5%自报
Wild Bench
65.3%自报
Arena Hard
43.1%自报
Language
MMLU-Pro
69.1%自报
Math
MATH
69.4%自报
MathVista
67.1%自报
Multimodal
MMMU
62.5%自报
Reasoning
HumanEval Plus
92.9%自报
ChartQAMasry et al. (2022)
87.4%自报
MBPP Plus
78.3%自报
GPQANYU + Cohere + Anthropic (2023)
46.1%自报
Vision
DocVQADocVQA (2020)
94.9%自报
AI2D
92.9%自报
AA 评测指数
(Artificial Analysis)Math 500(OpenAI (2024), subset of Hendrycks et al. MATH (2021))88.3
Mmlu Pro(TIGER-Lab (Univ. of Waterloo, Toronto, CMU, 2024))68.1
Gpqa(NYU + Cohere + Anthropic (2023))50.5
Ifbench(Google Research (2023))33.5
Aime(MAA (Mathematical Association of America))32.3
Tau2(Sierra + U Toronto + Vector Institute (2025))29.5
Scicode(UIUC + Argonne National Lab (2024))28.6
Livecodebench(UC Berkeley + MIT + Cornell (2024))27.5
Math Index(Artificial Analysis)27.0
Aime 25(MAA (Mathematical Association of America))27.0
Lcr(Artificial Analysis)20.3
Coding Index(Artificial Analysis)12.5
Intelligence Index(Artificial Analysis)8.2
Terminalbench Hard(Stanford × Laude Institute (2026))6.8
Tau Banking6.2
Terminalbench V2 15.6
Hle(Center for AI Safety + Scale AI (2025))4.3
Terminalbench V4 00.0
LLM Stats 分类评分
(LLM Stats (zeroeval))Image To Text90
Multimodal80
Vision80
Language70
Legal70
Math70
Finance70
General70
Healthcare70
Communication70
Reasoning60
Physics50
Biology50
Chemistry50
Chat40
Creativity40
Writing40
Factuality10
定价
输入价格$0.075 / 1M tokens
输出价格$0.2 / 1M tokens
混合价格(3:1)$0.106 / 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
3 个供应商
最便宜: DeepInfra最贵: DevPass (LLM Gateway)
供应商输入输出
1DeepInfra最便宜
$0
$0
2Mistral主要
$0.075
$0.2
3DevPass (LLM Gateway)
$0.1
$0.3
比较该模型在不同 API 供应商之间的定价。