Mistral Small 3.1

MistralMistral开源权重Apache 2.0 · 商用许可

描述

Mistral Small 3 is a 24B-parameter LLM licensed under Apache-2.0. It focuses on low-latency, high-efficiency instruction following, maintaining performance comparable to larger models. It provides quick, accurate responses for conversational agents, function calling, and domain-specific fine-tuning. Suitable for local inference when quantized, it rivals models 2–3× its size while using significantly fewer compute resources.

发布日期

2025-03-17

参数规模

24.0B

上下文长度

33K

支持模态

image, 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
代码能力榜	465	25.0	AA
通用能力榜	497	28.0	AA
多模态榜	108	42.0	LS
科学能力	503	27.0	AA

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Chat

MT-Bench

0.83 / 100自报

IFEvalGoogle Research (2023)

82.9%自报

Factuality

SimpleQA

10.4%自报

General

Arena Hard

87.6%自报

MMLU

80.6%自报

TriviaQA

80.5%自报

Wild Bench

52.2%自报

Language

MMLU-Pro

66.8%自报

Math

MATH

69.3%自报

Multimodal

MMMU

59.3%自报

Reasoning

HumanEvalOpenAI (2021)

88.4%自报

MBPP

0.75 / 100自报

GPQANYU + Cohere + Anthropic (2023)

46.0%自报

AA 评测指数

(Artificial Analysis)

Math 500(OpenAI (2024), subset of Hendrycks et al. MATH (2021))

70.7

Mmlu Pro(TIGER-Lab (Univ. of Waterloo, Toronto, CMU, 2024))

65.9

Gpqa(NYU + Cohere + Anthropic (2023))

45.4

Ifbench(Google Research (2023))

29.9

Scicode(UIUC + Argonne National Lab (2024))

27.8

Coding Index(Artificial Analysis)

26.3

Terminalbench V2 1

26.2

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

25.1

Lcr(Artificial Analysis)

22.3

Livecodebench(UC Berkeley + MIT + Cornell (2024))

21.2

Aime(MAA (Mathematical Association of America))

9.3

Terminalbench Hard(Stanford × Laude Institute (2026))

7.6

Tau Banking

7.4

Intelligence Index(Artificial Analysis)

7.1

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

4.3

Math Index(Artificial Analysis)

3.7

Aime 25(MAA (Mathematical Association of America))

3.7

Terminalbench V4 0

0.0

LLM Stats 分类评分

(LLM Stats (zeroeval))

Creativity

Writing

Chat

Instruction Following

Roleplay

Structured Output

Code

Language

Legal

Math

Reasoning

Finance

General

Healthcare

Communication

Physics

Biology

Chemistry

定价

输入价格$0.11 / 1M tokens

输出价格$0.17 / 1M tokens

混合价格(3:1)$0.125 / 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

2 个供应商

最便宜: DeepInfra最贵: Mistral

供应商输入输出

1DeepInfra最便宜

2Mistral主要

$0.11

$0.17

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