Mistral: Magistral Small 2506

Text input Text output
Author's Description

Magistral Small is a 24B parameter instruction-tuned model based on Mistral-Small-3.1 (2503), enhanced through supervised fine-tuning on traces from Magistral Medium and further refined via reinforcement learning. It is optimized for reasoning and supports a wide multilingual range, including over 20 languages.

Key Specifications
Cost
$$$$
Context
40K
Parameters
24B (Rumoured)
Released
Jun 10, 2025
Speed
Ability
Reliability
Supported Parameters

This model supports the following parameters:

Tool Choice Reasoning Include Reasoning Response Format Seed Top P Temperature Tools Structured Outputs Stop Max Tokens Frequency Penalty Presence Penalty
Features

This model supports the following features:

Tools Reasoning Response Format Structured Outputs
Performance Summary

Mistral: Magistral Small 2506 demonstrates a strong overall performance profile, particularly excelling in speed, where it typically performs among the fastest models, ranking in the top tier (64th percentile). It also offers competitive pricing, positioned in the 46th percentile. The model exhibits high reliability with a 95% success rate across benchmarks, indicating consistent technical stability. In terms of specific capabilities, Magistral Small shows notable strength in Coding, achieving 85.0% accuracy (60th percentile), and Email Classification, with an impressive 96.0% accuracy, albeit at a higher cost. However, its performance in Hallucinations (84.0% accuracy, 25th percentile), General Knowledge (71.0% accuracy, 21st percentile), Reasoning (44.0% accuracy, 28th percentile), and Ethics (88.0% accuracy, 22nd percentile) suggests areas for improvement, as these scores place it in the lower percentiles for accuracy. Instruction Following also presents a moderate challenge at 46.0% accuracy (44th percentile). While generally fast, its duration for Hallucinations and Instruction Following benchmarks was notably high.

Model Pricing

Current Pricing

Feature Price (per 1M tokens)
Prompt $0.5
Completion $1.5

Price History

Available Endpoints
Provider Endpoint Name Context Length Pricing (Input) Pricing (Output)
Mistral
Mistral | mistralai/magistral-small-2506 40K $0.5 / 1M tokens $1.5 / 1M tokens
Kluster
Kluster | mistralai/magistral-small-2506 40K $0.5 / 1M tokens $1.5 / 1M tokens
Enfer
Enfer | mistralai/magistral-small-2506 32K $0.5 / 1M tokens $1.5 / 1M tokens
Benchmark Results
Benchmark Category Reasoning Strategy Free Executions Accuracy Cost Duration
Other Models by mistralai