Magistral Small 1.1

Compact Transparent AI for Logic & Reasoning

What is Magistral Small 1.1?

Magistral Small 1.1 is a 24-billion parameter, open-source reasoning model from Mistral AI, focusing on precise, transparent, and stepwise outputs for technical, business, and regulated domains. Building on Mistral Small 3.1, it uses improved instruction tuning, reinforcement learning with reasoning traces from its Medium sibling, and enhanced formatting for interpretability. Outputs include clear chain-of-thought (“[THINK]...[/THINK]”) reasoning and LaTeX/Markdown support for technical tasks. It works locally on a single RTX 4090 or modern Mac, supporting efficient cloud and edge deployments.

Key Features of Magistral Small 1.1

Native Reasoning & Traceability

Produces auditable, stepwise logic inside “[THINK]” tokens, clarifying how the model solves complex technical, business, and legal problems.

Improved Tone, Conciseness, and Robustness

Shorter, more relevant responses on easy queries; better formatting for LaTeX and Markdown; lower risk of infinite text or hallucination.

Multilingual Capabilities

Handles logic and reasoning assignments in English, French, Spanish, German, Italian, Arabic, Russian, and Chinese.

Fast Local & Edge Inference

Runs efficiently on consumer hardware (single GPU or MacBook), ideal for privacy-oriented or regulated business scenarios.

Open Source & Fine-Tuned

Released with Apache 2.0 license and instruction-tuned on real enterprise traces for transparent, iterative improvement.

Use Cases of Magistral Small 1.1

Brings step-by-step logic for legal, compliance, risk, and regulatory reviews, ensuring all conclusions are fully traceable and auditable.

Performs complex calculations, data-backed delivery/supply chain predictions, and multi-factor business scenario simulations.

Enhances system design, code review, planning, and debugging via readable reasoning traces and agentic integration with external tools.

Delivers logic-based output for chatbots, programmatic interfaces, and multi-language support in business, government, and healthcare.

Powers technical writing and creative storytelling with explicit, interpretable chains of logic that aid in learning or analysis.

Magistral Small 1.1

vs

Comparable Models

Feature	Magistral Small 1.1	Magistral Medium	Mistral Small 3.1	Claude Sonnet 3.7
Reasoning Logic	Fully traceable	Advanced, larger	Standard CoT	Standard
Auditability	High, stepwise	High, larger	Basic	Basic
Speed	194 tokens/sec	10x base models	~150 tokens/sec	~80 tokens/sec
Languages	8+ (high-fidelity)	8+	21+	30+
Platform	Local/cloud/edge	Cloud/API	Local/cloud/edge	Cloud
License	Apache 2.0 open	Enterprise	Apache 2.0 open	Proprietary