Mistral Small 4 (Reasoning) logo
Mistral
•

Open weights model

•

Released March 2026

Análise de inteligência, desempenho e preço do Mistral Small 4 (Reasoning)

Experimente Benchmarks de provedores de API

Resumo do modelo

Inteligência

20
Artificial Analysis Intelligence Index
4 de 4 unidades para Inteligência.

Velocidade

160.3
Tokens de saída por segundo
4 de 4 unidades para Velocidade.

Preço de entrada

US$ 0,15
USD por 1M de tokens
2 de 4 unidades para Preço de entrada.

Preço de saída

US$ 0,60
USD por 1M de tokens
3 de 4 unidades para Preço de saída.

Verbosidade

53M
Tokens de saída do Intelligence Index
2 de 4 unidades para Verbosidade.

Mistral Small 4 (Reasoning) está entre os modelos líderes em inteligência e tem preço razoável em comparação com outros modelos de pesos abertos de tamanho semelhante. Também é especialmente rápido e bastante conciso. O modelo aceita entrada em texto e imagem, gera saída em texto e tem uma janela de contexto de 256k tokens.

Mistral Small 4 (Reasoning) alcança 20 pontos no Artificial Analysis Intelligence Index, ficando bem acima da média entre os modelos comparáveis (mediana: 9). Na avaliação do Intelligence Index, gerou 53M tokens, um resultado bastante conciso em comparação com a mediana de 53M.

O preço do Mistral Small 4 (Reasoning) é de $0.15 por 1M de tokens de entrada (preço moderado, mediana: $0.15) e $0.60 por 1M de tokens de saída (preço um pouco alto, mediana: $0.45). No total, a avaliação do Mistral Small 4 (Reasoning) no Intelligence Index custou $87.94.

Com 160 tokens por segundo, Mistral Small 4 (Reasoning) é especialmente rápido (93).

RaciocínioSim

Esta página mostra a versão com raciocínio deste modelo.

Também pode existir uma variante sem raciocínio.

Modalidade de entrada

Compatível com: texto e imagem

Modalidade de saída

Compatível com: texto

Janela de contexto256k
~384 páginas A4 em fonte Arial tamanho 12
Parâmetros totais119B
Parâmetros ativos6.5B
Número de parâmetros ativos por token durante a inferência
LicençaApache 2.0
Pesos do modeloHugging Face

As métricas são comparadas com modelos da mesma classe:

  • Modelos sem raciocínio → comparados apenas com outros modelos sem raciocínio
  • Modelos de raciocínio → comparados com modelos com e sem raciocínio
  • Modelos de pesos abertos → comparados apenas com outros modelos de pesos abertos da mesma classe de tamanho:
    • Muito pequeno: ≤4B parâmetros
    • Pequeno: 4B–40B parâmetros
    • Médio: 40B–150B parâmetros
    • Grande: >150B parâmetros
  • Modelos proprietários → comparados com modelos proprietários e de pesos abertos da mesma faixa de preço, usando uma proporção combinada de 3:1 entre os preços de entrada e saída:
    • <$0.15 por 1M de tokens
    • $0.15–$1 por 1M de tokens
    • >$1 por 1M de tokens

Destaques

Inteligência

Artificial Analysis Intelligence Index · Higher is better

Velocidade

Output tokens per second · Higher is better

Custo por tarefa

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

Inteligência

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index v4.1 incorporates 9 evaluations: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR
Reasoning models are indicated by a lightbulb icon

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Artificial Analysis Intelligence Index by Open Weights / Proprietary

Artificial Analysis Intelligence Index v4.1 incorporates 9 evaluations: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR
Reasoning models are indicated by a lightbulb icon

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Indicates whether the model weights are available. Models are labelled as 'Commercial Use Restricted' if the weights are available but commercial use is limited (typically requires obtaining a paid license).

Avaliações

Intelligence Evaluations

Intelligence evaluations measured independently by Artificial Analysis · Higher is better
GDPval-AA v2

Agentic real-world work tasks, (Elo-500)/2000

𝜏³-Banking

Agentic tool use

Terminal-Bench v2.1

Agentic coding & terminal use

SciCode

Coding

Humanity's Last Exam

Reasoning & knowledge

GPQA Diamond

Scientific reasoning

CritPt

Physics reasoning

AA-Omniscience Accuracy

Knowledge

AA-Omniscience Non-Hallucination Rate

1 - hallucination rate

AA-LCR

Long context reasoning

AA-Briefcase

Agentic knowledge work, Elo

AutomationBench-AA

Agentic SaaS workflows

Harvey LAB-AA

Legal agentic work, criterion pass rate

EnterpriseOps-Gym-AANew

Agentic business operations

IFBench

Instruction following

APEX-Agents-AA

Long-horizon agentic tasks

ITBench-AA

Kubernetes incident root-cause analysis

MMMU-Pro

Visual reasoning

Reasoning models are indicated by a lightbulb icon

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

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

AA-Omniscience

AA-Omniscience Index

AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct.
Reasoning models are indicated by a lightbulb icon

AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct.

Openness Index

Artificial Analysis Openness Index: Score

Openness Index assesses model openness on a 0 to 100 normalized scale (higher is more open)
Reasoning models are indicated by a lightbulb icon

Comparações do Intelligence Index

Intelligence Index vs. Cost per Intelligence Index Task

Artificial Analysis Intelligence Index · Weighted average cost (USD) per Artificial Analysis Intelligence Index task
Most attractive quadrant
Pareto line
Reasoning models are indicated by a lightbulb icon

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.

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Uso de tokens

Output Tokens per Intelligence Index Task

Weighted average number of output tokens used to run one task in the Artificial Analysis Intelligence Index
Reasoning models are indicated by a lightbulb icon

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).

Custo

Cost per Intelligence Index Task

Weighted average cost (USD) per Artificial Analysis Intelligence Index task, segmented by token type. Lower is better
Reasoning models are indicated by a lightbulb icon

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.

Cost to Run Artificial Analysis Intelligence Index

Cost (USD) to run all evaluations in the Artificial Analysis Intelligence Index
Reasoning models are indicated by a lightbulb icon

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).

Pricing: Cache Hit, Input, and Output

Price (USD per M Tokens)
Reasoning models are indicated by a lightbulb icon

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.

Janela de contexto

Context Window

Context window: tokens limit · Higher is better
Reasoning models are indicated by a lightbulb icon

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).

Velocidade

Medida pela velocidade de saída (tokens por segundo)

Output Speed

Output tokens per second · Higher is better
Reasoning models are indicated by a lightbulb icon

Tokens per second received while the model is generating tokens (ie. after first chunk has been received from the API for models which support streaming).

Figures represent performance of the model's first-party API (e.g. OpenAI for o1) or the median across providers where a first-party API is not available (e.g. Meta's Llama models).

Time per Intelligence Index Task

Weighted average decode time (minutes) per task; excludes TTFT and overhead time · Lower is better
Reasoning models are indicated by a lightbulb icon

The weighted average time (seconds) per Artificial Analysis Intelligence Index task. This is calculated by dividing output tokens per task by output speed, weighted by the relative weights of each benchmark in the Intelligence Index.

Latência

Medida pelo tempo (segundos) até o primeiro token

Latency: Time To First Answer Token

Seconds to first answer token received · Accounts for reasoning model 'thinking' time
Reasoning models are indicated by a lightbulb icon

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.

Tempo de resposta de ponta a ponta

Seconds to output 500 tokens, calculated based on time to first token, 'thinking' time for reasoning models, and output speed

End-to-End Response Time

Seconds to output 500 tokens, including reasoning model 'thinking' time · Lower is better
Reasoning models are indicated by a lightbulb icon

Seconds to receive a 500 token response. Key components:

  • Input time: Time to receive the first response token
  • Thinking time (only for reasoning models): Time reasoning models spend outputting tokens to reason prior to providing an answer. Amount of tokens based on the average reasoning tokens across a diverse set of 60 prompts (methodology details).
  • Answer time: Time to generate 500 output tokens, based on output speed

Figures represent performance of the model's first-party API (e.g. OpenAI for o1) or the median across providers where a first-party API is not available (e.g. Meta's Llama models).

Tamanho do modelo (somente modelos de pesos abertos)

Model Size: Total and Active Parameters

Comparison between total model parameters and parameters active during inference
Reasoning models are indicated by a lightbulb icon

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.

Perguntas frequentes

Perguntas comuns sobre o Mistral Small 4 (Reasoning)

O Mistral Small 4 (Reasoning) foi lançado em 16 de março de 2026.

Mistral criou o Mistral Small 4 (Reasoning).

Mistral Small 4 (Reasoning) alcança 20 pontos no Artificial Analysis Intelligence Index, ficando bem acima da média entre outros modelos de pesos abertos de tamanho semelhante (mediana: 9).

Mistral Small 4 (Reasoning) gera a saída a 160.3 tokens por segundo (com base na API de Mistral), um resultado bem acima da média em comparação com outros modelos de pesos abertos de tamanho semelhante (mediana: 93.2 t/s).

Mistral Small 4 (Reasoning) tem um tempo até o primeiro token (TTFT) de 1.05s (com base na API de Mistral), um resultado muito competitivo em comparação com outros modelos de pesos abertos de tamanho semelhante (mediana: 1.80s).

Mistral Small 4 (Reasoning) custa $0.15 por 1M de tokens de entrada (muito competitivo, mediana: $0.40) e $0.60 por 1M de tokens de saída (melhor que a média, mediana: $0.71), com base na API de Mistral.

Mistral Small 4 (Reasoning) custa $0.15 por 1M de tokens de entrada e $0.60 por 1M de tokens de saída (com base na API de Mistral). Considerando uma tarifa combinada (proporção de 7:2:1 entre acerto do cache, entrada e saída), o preço é de $0.20 por 1M de tokens. O preço pode variar conforme o provedor. Comparar preços dos provedores

Na avaliação do Intelligence Index, Mistral Small 4 (Reasoning) gerou 53M tokens de saída, um resultado melhor que a média em comparação com outros modelos de pesos abertos de tamanho semelhante (mediana: 53M).

Sim, o Mistral Small 4 (Reasoning) é um modelo de raciocínio. Ele usa reflexão prolongada ou raciocínio em cadeia para resolver problemas complexos antes de fornecer uma resposta.

Mistral Small 4 (Reasoning) aceita entrada em texto e imagem.

Mistral Small 4 (Reasoning) permite saída em texto.

Sim, o Mistral Small 4 (Reasoning) aceita imagens como entrada e pode analisá-las, descrevê-las e responder a perguntas sobre elas.

Sim, o Mistral Small 4 (Reasoning) é multimodal. Ele consegue processar entrada em texto e imagem e gerar saída em texto.

Mistral Small 4 (Reasoning) tem uma janela de contexto de 260k tokens. Isso determina quanto texto e histórico da conversa o modelo consegue processar em uma única solicitação.

Sim, o Mistral Small 4 (Reasoning) é um modelo de pesos abertos. Os pesos estão disponíveis publicamente e podem ser baixados para auto-hospedagem.

Mistral Small 4 (Reasoning) tem 119 bilhões parâmetros (6,5 bilhões ativos).

Mistral Small 4 (Reasoning) é um modelo de mistura de especialistas (MoE) com 119 bilhões parâmetros no total, mas apenas 6,5 bilhões parâmetros ativos são usados durante a inferência.

Mistral Small 4 (Reasoning) foi lançado sob a licença Apache 2.0. Essa licença permite o uso comercial. Ver licença

Mistral Small 4 (Reasoning) alcança 20 pontos no Artificial Analysis Intelligence Index. Esse benchmark composto avalia os modelos em raciocínio, conhecimento, matemática e programação.

Sim, o Mistral Small 4 (Reasoning) está disponível por API em 1 provedor. Comparar provedores de API

Mistral Small 4 (Reasoning) está disponível em 1 provedor de API. Comparar provedores