AI Radar

Mistral Small 3

Available · Language · Milestone

Mistral Small 3 is a 24B-parameter language model from Mistral AI, released with pretrained and instruction-tuned open weights under Apache 2.0 and offered via API on la Plateforme and through partners. Mistral AI describes it as latency-optimised, with far fewer layers than competing models, and says it was trained without reinforcement learning or synthetic data. [1] [2] Primary source

Timeline of Mistral Small 3 →

Claims and evidence

  • Released Primary source[1]page date; available on la Plateforme as mistral-small-2501 [3]January 30, 2025 [4]release date
  • Deprecated Primary source[5]Deprecated table, Mistral Small 3.0 (mistral-small-2501)
  • Retired Primary source[5]Deprecated table, Mistral Small 3.0 (mistral-small-2501) [4]Retirement date
  • Status AvailableApache 2.0 weights are still downloadable from Mistral AI's Hugging Face organisation; the API model mistral-small-2501 was retired on 2025-11-30. Primary source[2] [6] [7]Model Lifecycle: Mistral Small 3.0 Instruct and Base, status Active
  • Replaced by Mistral Small 4Replacement named for the API model when checked on 2026-10-01. Primary source[5]Deprecated table, Mistral Small 3.0, Alternative [4]Replacement
  • Successor of Mistral Small v24.09 EditorialEditorial link along the Mistral Small line: the announcement names no predecessor, and Mistral Small v24.09 (Mistral Small 2.0 in the docs) was the previous Mistral Small release. Primary source[1]
  • Change · Licensing Released under Apache 2.0; Mistral AI said it would move away from MRL-licensed models for its general-purpose models.Compared with Mistral Small v24.09 Primary source[1]
  • Change · Size Has 24B parameters, published as a pretrained and an instruction-tuned checkpoint.Compared with Mistral Small v24.09 Primary source[1] [2]
  • Change · Architecture Mistral AI says it has far fewer layers than competing models, which reduces the time per forward pass.Compared with Mistral Small v24.09 Primary source[1]
  • Input text Primary source[4]Modalities
  • Output text Primary source[4]Modalities
  • Open weights Yes Primary source[1]pretrained and instruction-tuned checkpoints under Apache 2.0 [2] [6]
  • Context window 32k tokens Primary source[4]Context [2]Key Features: Context Window
  • Parameters 24B Primary source[1]first paragraph [4]Weights table, Parameters (B)
  • Access API, Open-weights download, cloud partnerAt launch on la Plateforme, on Hugging Face, Ollama and Kaggle, and through Together AI, Fireworks AI and IBM watsonx. Primary source[1]Using Mistral Small 3 on your preferred tech stack

Lineage

Predecessors

Successors

Based on

Not derived from another model.

Variants and derived

None recorded.

Siblings

None recorded.

All ancestors

All descendants

Variants

No variants recorded in this record.

Related AI Radar coverage

AI Radar coverage starts in June 2026; no coverage linked yet.

All model releases from Mistral AI on AI Radar →