Magistral Small 24B

Good fit for

People who want to evaluate a reasoning-focused model for multi-step text tasks.

Why consider it

The runtime entry documents multilingual reasoning, code-related workflows and creative writing.

Main limitations

Although 128K context is listed, the documentation warns of degradation beyond 40K and recommends a 40K maximum. Reasoning traces do not guarantee accuracy, regulatory compliance or suitability for legal, financial or medical decisions. No specialized professional-advice promise is copied into the listing.

Cost and access

Local package without a hosted subscription or per-token inference charge. Hardware, electricity and setup remain separate. License: Apache 2.0.

Getting started

Install a compatible Ollama version, review the model license, then run: ollama run magistral:24b. API access means the local runtime endpoint, not an included hosted API.