Mistral Small 4 (Reasoning) 在 Artificial Analysis Intelligence Index 上的得分为 4，在同类模型中低于平均水平（中位数：4）。在 Intelligence Index 评测中，它生成了 23.9M 个 token；与 63.8M 的中位数相比，其回答非常简洁。

Mistral Small 4 (Reasoning) 每 100 万输入 token 的价格为 $0.15（很有竞争力，中位数：$0.30），每 100 万输出 token 的价格为 $0.60（中等，中位数：$0.80）。在 Intelligence Index 上评测 Mistral Small 4 (Reasoning) 的平均每项任务成本为 $0.0399。

Mistral Small 4 (Reasoning) 完成一项 Intelligence Index 任务需 1.7 分钟，与同类模型 5.3 分钟的中位数相比明显较短（越低越好）。

输入模态	支持：文本和图像
输出模态	支持：文本
上下文窗口	256k 约 384 页 A4 纸（12 号 Arial 字体）
总参数量	119B
活跃参数量	6.5B 推理期间每个 token 启用的参数数量
许可证	Apache 2.0
模型权重	Hugging Face

指标与同类别模型进行比较：

非推理模型 → 仅与其他非推理模型比较
推理模型 → 同时与推理和非推理模型比较
开放权重模型 → 仅与规模类别相同的其他开放权重模型比较：

微型：≤4B 个参数
小型：4B–40B 个参数
中型：40B–150B 个参数
大型：>150B 个参数

专有模型 → 与价格区间相同的专有模型和开放权重模型比较，采用输入/输出价格 3:1 的混合比例：

每 100 万 token <$0.15
每 100 万 token $0.15–$1
每 100 万 token >$1

智能

Artificial Analysis Intelligence Index · Higher is better

No publicly available API

每任务耗时

Inference time (minutes) per Intelligence Index task · Lower is better

No publicly available API

每任务成本

Weighted average cost (USD) per Intelligence Index task · Lower is better

No publicly available API

智能

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index v5.0.1 incorporates 9 evaluations: AA-Briefcase v1.1, AutomationBench-AA, Terminal-Bench 4.0, FrontierCode-AA 2.0, AA-Omniscience v1.1, GDP.pdf, HLE-Diamond, CritPt v2 (Hard Subset), and Terminal-Bench-Science 0.1

Artificial Analysis Intelligence Index v5.0.1 includes: AA-Briefcase v1.1, AutomationBench-AA, Terminal-Bench 4.0, FrontierCode-AA 2.0, AA-Omniscience v1.1, GDP.pdf, HLE-Diamond, CritPt v2 (Hard Subset), Terminal-Bench-Science 0.1. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Artificial Analysis Intelligence Index：开放权重与专有模型

Indicates whether the model weights are available. Models are labelled as 'Commercial Use Restricted' if commercial use is limited by conditions, and as 'Non-commercial' if the license prohibits commercial use.

Intelligence Index 与每项任务的成本

Artificial Analysis Intelligence Index · Weighted average cost (USD) per Artificial Analysis Intelligence Index task

Most attractive quadrant

Pareto line

Weighted average cost per Intelligence Index task. Each evaluation’s cost is calculated from input, cache hit, cache write, reasoning, and answer token prices, divided by task count, and weighted by its Intelligence Index weight.

能力指数

衡量模型在特定能力与行业中的表现

Finance & Accounting Index

Incorporates 5 evaluations: AA-Briefcase v1.1, AA-Omniscience v1.1, HLE-Diamond, AutomationBench-AA, GDP.pdf · Higher is better

Strategy & Ops Index

Incorporates 4 evaluations: AutomationBench-AA, AA-Briefcase v1.1, AA-Omniscience v1.1, GDP.pdf · Higher is better

Legal Index

Incorporates 5 evaluations: AA-Omniscience v1.1, AA-Briefcase v1.1, HLE-Diamond, GDP.pdf, AutomationBench-AA · Higher is better

Healthcare & Medical Index

Incorporates 5 evaluations: AA-Omniscience v1.1, AA-Briefcase v1.1, MLCR-AA, AutomationBench-AA, HLE-Diamond · Higher is better

Engineering Index

Incorporates 7 evaluations: AA-Omniscience v1.1, CritPt v2 (Hard Subset), Terminal-Bench-Science 0.1, HLE-Diamond, Terminal-Bench 4.0, FrontierCode-AA 2.0, AA-Briefcase v1.1 · Higher is better

Economics Index

Incorporates 4 evaluations: AA-Omniscience v1.1, AA-Briefcase v1.1, HLE-Diamond, GDP.pdf · Higher is better

基准测试

Evaluations

Intelligence evaluations measured independently by Artificial Analysis · Higher is better

AA-Briefcase v1.1

Agentic knowledge work, Elo

AA-Omniscience v1.1 Accuracy

Knowledge

AA-Omniscience v1.1 Non-Hallucination Rate

1 - hallucination rate

AutomationBench-AA

Agentic SaaS workflows

Terminal-Bench-Science 0.1

Agentic scientific research workflows in a terminal

FrontierCode-AA 2.0

Agentic software engineering in open-source repositories

HLE-Diamond

Reasoning & knowledge

GDP.pdf

Professional document reasoning, All-pass

Terminal-Bench 4.0

Agentic coding & terminal use

CritPt v2 (Hard Subset)

Physics reasoning

AA-AnalystAgent

Quantitative analysis on spreadsheets & documents

GDPval-AA v2.1

Agentic real-world work tasks, Elo

MLCR-AA

Medical long context reasoning

MMMU-Pro

Visual reasoning

While model intelligence generally translates across use cases, specific evaluations may be more relevant for certain use cases.

Openness Index

Artificial Analysis Openness Index：得分

Openness Index 以 0 至 100 的标准化评分衡量模型的开放程度（分数越高，越开放）

成本

每项 Intelligence Index 任务的成本

Weighted average cost (USD) per Intelligence Index task · Segmented by token type · Lower is better

运行 Artificial Analysis Intelligence Index 的成本

Cost (USD) to run all evaluations in the Artificial Analysis Intelligence Index

The cost to run the evaluations in the Artificial Analysis Intelligence Index, calculated using the model's input, cache hit, cache write, reasoning, and answer token prices and the number of tokens used across evaluations (excluding repeats).

价格：缓存命中、输入和输出

Price (USD per M Tokens)

Price per token for cached prompts (previously processed), typically offering a significant discount compared to regular input price, represented as USD per million tokens. The values shown here are the cache hit price; cache write and cache storage are billed separately and vary by provider — see "Cache pricing by provider" for detail.

性能

按每任务耗时（分钟）、输出速度（每秒 token 数）、首个答案 Token 延迟和首 Token 延迟（秒）衡量

Intelligence Index 每项任务耗时

Output tokens divided by end-to-end output speed (minutes per Intelligence Index task) · Speed at the model's own endpoint or the median across providers · Lower is better

The weighted average time (minutes) per Artificial Analysis Intelligence Index task: output tokens per task divided by output speed, weighted by each benchmark's weight in the Intelligence Index.

Speed and latency metrics average each endpoint's runs over the past 5 days, and the error rate pools them. We run the benchmark every 6 hours, and endpoints of deprecated models every 18 hours.

输出速度

Output tokens divided by total request time, including time to first token · Higher is better

Output tokens, including reasoning tokens, received per second over a request's full duration, from sending the request to receiving the final token. The headline value averages each Performance v3 workload over the past 5 days, then averages the workloads.

首个答案 Token 延迟

收到首个答案 token 的秒数 · 包含推理模型的“思考”时间

Time to first answer token received, in seconds, after API request sent. For reasoning models, this includes the 'thinking' time of the model before providing an answer. For models which do not support streaming, this represents time to receive the completion.

Token 使用量

Intelligence Index 每项任务的输出 Token 数

运行 Artificial Analysis Intelligence Index 单项任务所用输出 token 数的加权平均值

The number of tokens required per Intelligence Index task. This is calculated by multiplying the output tokens per eval by the relative weights of each benchmark in the Intelligence Index, then dividing by task count (excluding repeats).

上下文窗口

Context window: tokens limit · Higher is better

Larger context windows are relevant to RAG (Retrieval Augmented Generation) LLM workflows which typically involve reasoning and information retrieval of large amounts of data.

Maximum number of combined input & output tokens. Output tokens commonly have a significantly lower limit (varied by model).

模型规模（仅开放权重模型）

模型规模：总参数量与活跃参数量

Comparison between total model parameters and parameters active during inference (billions)

The total number of trainable weights and biases in the model, expressed in billions. These parameters are learned during training and determine the model's ability to process and generate responses.

The number of parameters actually executed during each inference forward pass, expressed in billions. For Mixture of Experts (MoE) models, a routing mechanism selects a subset of experts per token, resulting in fewer active than total parameters. Dense models use all parameters, so active equals total.

Mistral 的更多发布

常见问题

关于 Mistral Small 4 的常见问题

Mistral Small 4 发布于 2026年3月16日。

Mistral Small 4 由 Mistral 开发。

Mistral Small 4 共有 2 个思考强度等级：reasoning和non-reasoning。Mistral Small 4 在 Artificial Analysis Intelligence Index 上的得分为 4（reasoning）。Mistral Small 4 (Reasoning) 在其他规模相近的开放权重模型中低于平均水平（中位数：4）。

Mistral Small 4 完成一项 Intelligence Index 任务需 1.7 分钟（reasoning）（基于 Mistral 的 API，越低越好）。Mistral Small 4 (Reasoning) 与其他规模相近的开放权重模型相比属于最快之列（中位数：5.3 分钟）。

Mistral Small 4 的首个答案 Token 延迟为 12.91 秒（reasoning）和1.58 秒（non-reasoning）（基于 Mistral 的 API）。Mistral Small 4 (Reasoning) 的首个答案 Token 延迟与其他规模相近的开放权重模型相比略高于平均水平（中位数：8.59 秒）。

Mistral Small 4 每 100 万输入 token 的价格为 $0.15（非常有竞争力，中位数：$0.30），每 100 万输出 token 的价格为 $0.60（优于平均水平，中位数：$0.80）；基于 Mistral 的 API。

Mistral Small 4 每 100 万输入 token 的价格为 $0.15，每 100 万输出 token 的价格为 $0.60（基于 Mistral 的 API）。按缓存命中/输入/输出为 7:2:1 的比例计算，混合价格为每 100 万 token $0.10。价格可能因服务商而异。比较服务商价格

在 Intelligence Index 评测中，Mistral Small 4 生成了 23.9M 个输出 token（reasoning）。Mistral Small 4 (Reasoning) 的输出 token 数与其他规模相近的开放权重模型相比非常有竞争力（中位数：63.8M）。

是，Mistral Small 4 (Reasoning) 是推理模型。它会在给出答案之前，通过扩展思考或思维链推理来解决复杂问题。否，Mistral Small 4 (Non-reasoning) 不是推理模型。它不进行扩展的思维链推理，而是直接作答。

Mistral Small 4 支持文本和图像输入。

Mistral Small 4 支持文本输出。

是，Mistral Small 4 支持图像输入，可以分析、描述图像并回答有关图像的问题。

是，Mistral Small 4 是多模态模型，可以处理文本和图像输入并生成文本输出。

Mistral Small 4 的上下文窗口为 256k 个 token。这决定了模型在单次请求中可以处理多少文本和对话历史。

是，Mistral Small 4 是开放权重模型，其模型权重已公开，可供下载并自行托管。

Mistral Small 4 有 119B 参数（启用 6.5B）。

Mistral Small 4 是混合专家（MoE）模型，总参数量为 119B，但推理期间仅启用 6.5B。

Mistral Small 4 基于 Apache 2.0 许可证发布，该许可证允许商业使用。查看许可证

Mistral Small 4 在 Artificial Analysis Intelligence Index 上的得分为 4（reasoning）。这项综合基准测试评估模型的知识工作、编程、科学和通用知识能力。

是，可通过 1 家服务商的 API 使用 Mistral Small 4。比较 API 服务商

可通过 1 家 API 服务商使用 Mistral Small 4。比较服务商

								权重
Mistral Small 4 (Reasoning)	Mistral	4	119B 推理时启用 6.5B 个参数	256k	$0.10	1.7	142	🤗
Mistral Small 4 (Non-reasoning)	Mistral	-	119B 推理时启用 6.5B 个参数	256k	$0.10	-	120	🤗

Mistral Small 4：智能、性能与价格分析

模型摘要

技术规格

此类别中有 67 个模型

智能

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index：开放权重与专有模型

Artificial Analysis Intelligence Index

Open Weights

Intelligence Index 与每项任务的成本

Cost per Intelligence Index Task

Artificial Analysis Intelligence Index

能力指数

Evaluations

Intelligence Evaluation Relevance

Artificial Analysis Intelligence Index

Openness Index

Artificial Analysis Openness Index：得分

成本

每项 Intelligence Index 任务的成本

Cost per Intelligence Index Task

运行 Artificial Analysis Intelligence Index 的成本

Cost to Run Artificial Analysis Intelligence Index

价格：缓存命中、输入和输出

Cache Hit

性能

Intelligence Index 每项任务耗时

Time per Intelligence Index Task

Time Period

输出速度

End-to-End Output Speed

首个答案 Token 延迟

Time to First Answer Token

Token 使用量

Intelligence Index 每项任务的输出 Token 数

Output Tokens per Intelligence Index Task

上下文窗口

上下文窗口

Context Window for RAG

Context Window

模型规模（仅开放权重模型）

模型规模：总参数量与活跃参数量

Total Parameters

Active Parameters at Inference Time

更多详情

Mistral 的更多发布

常见问题

Mistral Small 4 何时发布？

谁开发了 Mistral Small 4？

Mistral Small 4 有多智能？

Mistral Small 4 有多快？

Mistral Small 4 的延迟是多少？

Mistral Small 4 的费用是多少？

Mistral Small 4 API 的价格是多少？

Mistral Small 4 的回答有多冗长？

Mistral Small 4 是推理模型吗？

Mistral Small 4 支持哪些输入模态？

Mistral Small 4 支持哪些输出模态？

Mistral Small 4 能处理图像吗？

Mistral Small 4 是多模态模型吗？

Mistral Small 4 的上下文窗口有多大？

Mistral Small 4 是开放权重模型吗？

Mistral Small 4 有多少个参数？

Mistral Small 4 有多少活跃参数？

Mistral Small 4 使用什么许可证？

Mistral Small 4 在基准测试中的表现如何？

可以通过 API 使用 Mistral Small 4 吗？

可以在哪里使用 Mistral Small 4？