Summary
Mistral AI released Mistral Small 3, a 24B-parameter language model, under Apache 2.0 on January 30, 2025. Apache 2.0 permits commercial use, modification, and redistribution subject to its conditions, including preservation of relevant notices; it does not provide a blanket exemption from attribution requirements.
Editorial correction, October 26, 2026
Apache 2.0 section 4 requires a copy of the license for recipients, notices of modified files, retention of relevant source notices, and applicable NOTICE attribution in redistributed derivative works. The original body's unconditional no-attribution wording is withdrawn. This describes the license text rather than promising that any particular distribution satisfies it.
Revision record with the prior summary and metadata.
The original narrative below is preserved as published; use the revised summary and this note for the corrected account.
What Happened
Released on January 30, 2025, Mistral Small 3 is a 24B dense transformer with a 32K token context window and integrated support for structured function calling. The model achieved over 81% on the MMLU benchmark and demonstrated strong performance on coding and instruction-following tasks relative to its parameter count. Mistral positioned it explicitly as a deployment-optimized model for developers running inference locally or on constrained cloud infrastructure.
The Apache 2.0 license — one of the most permissive available — allowed unrestricted commercial use, modification, and redistribution without attribution requirements. This distinguished it from models using custom licenses with carve-outs for distillation or commercial scale. Model weights were published on Hugging Face.
The 24B scale sits in a practical range for single-GPU or two-GPU inference on consumer hardware, making it accessible for teams building latency-sensitive applications or working in regulated environments where data cannot leave local infrastructure.
Why It Matters
Mistral Small 3 reinforced Mistral's strategic positioning as a European open-source counterweight to US frontier labs. By pairing genuine functional capability with a fully permissive license, it offered a credible alternative to closed API products for developers who prioritized cost control and data sovereignty.
The release also contributed to a growing body of evidence that the 20-30B parameter range could serve most enterprise use cases at a fraction of the cost of 100B+ models. As the open-weight ecosystem matured, Mistral Small 3 exemplified the "good enough open" tier — models that didn't claim frontier performance but were freely deployable and practically capable.
§ How to read the metadata
- Landmark
- Fundamentally alters the trajectory; 2–5 per year.
- Major
- Meaningfully shifts the landscape; 2–4 per month.
- Notable
- Worth documenting; significance can be upgraded later.
- Confidence
- High = primary sources corroborate. Medium = credible secondary only. Low = provisional. Disputed = credible sources disagree.
- Contestation
- Uncontested = no formal challenge. Contested = at least one challenge open. Superseded = replaced by a later entry. Unresolved = dispute still open.