AIbase
EN
Home
AI Models Finder
Mistral Small 3.2
M

Mistral Small 3.2

Latest small language model from Mistral AI with approximately 22B parameters, supporting 128K context window. Sets new performance benchmarks in the sub-70B model category with enhanced reasoning capabilities and improved multilingual support. Optimized for deployment efficiency, offers excellent performance-to-size ratio, suitable for enterprise applications balancing performance and resource consumption.
Intelligence(Medium)
Speed(Relatively Fast)
Input Supported Modalities
No
Is Reasoning Model
128,000
Context Window
32,000
Maximum Output Tokens
2023-10-01
Knowledge Cutoff
Go Compare

Pricing

￥0.5 /M tokens
Input
￥1.01 /M tokens
Output
￥1.08 /M tokens
Blended Price

Quick Simple Comparison

Input

Output

Devstral
￥0.1
Devstral Medium
￥0.4
Devstral Small (May '25)
￥0.1

Basic Parameters

Mistral Small 3.2Technical Parameters
Parameter Count
24,000.0M
Context Length
128.00k tokens
Training Data Cutoff
2023-10-01
Open Source Category
Open Weights (Permissive License)
Multimodal Support
Text Only
Throughput
134
Release Date
2025-06-20
Response Speed
183.29,317 tokens/s

Benchmark Scores

Below is the performance of Mistral Small 3.2 in various standard benchmark tests. These tests evaluate the model's capabilities in different tasks and domains.
Intelligence Index
42.29
Large Language Model Intelligence Level
Coding Index
26.95
Indicator of AI model performance on coding tasks
Math Index
60.3
Capability indicator in solving mathematical problems, mathematical reasoning, or performing math-related tasks
MMLU Pro
68.1
Massive Multitask Multimodal Understanding - Testing understanding of text, images, audio, and video
GPQA
50.5
Graduate Physics Questions Assessment - Testing advanced physics knowledge with diamond science-level questions
HLE
4.3
The model's comprehensive average score on the Hugging Face Open LLM Leaderboard
LiveCodeBench
27.5
Specific evaluation focused on assessing large language models' ability in real-world code writing and solving programming competition problems
SciCode
26.4
The model's capability in code generation for scientific computing or specific scientific domains
HumanEval
85.2
Score achieved by the AI model on the specific HumanEval benchmark test set
Math 500 Score
88.3
Score on the first 500 larger, more well-known mathematical benchmark tests
AIME Score
32.3
An indicator measuring an AI model's ability to solve high-difficulty mathematical competition problems (specifically AIME level)
AIbase
Empowering the Future, Your AI Solution Knowledge Base
English简体中文繁體中文にほんご
© 2025AIbase