Mistral Small

weights

open(safetensors on HF mistralai/Mistral-Small-4-119B-2603, ungated

data

closed

code

partial(inference and serving only

license

Apache-2.0(OSI)

Mistral Small 4's weights are downloadable without a gate under Apache-2.0, an open-source license, but Mistral publishes neither the training data nor the training code. That makes it an open-weights release with a closed recipe.

• https://docs.mistral.ai/models/mistral-small-4-0-26-03 recorded 2026-10-13

  Mistral's model page: "March 16, 2026", "Apache 2.0", "v 26.03", "Mistral Small 4", "Our powerful hybrid model unifying instruct, reasoning, and coding capabilities in a single model", API identifier "mistral-small-2603", context 256k, priced at $0.15 per million input tokens and $0.6 per million output tokens.

• https://huggingface.co/api/models?author=mistralai&search=Mistral-Small-4&full=true recorded 2026-10-13

  Mistral's Hub listing for Mistral Small 4: Mistral-Small-4-119B-2603, Mistral-Small-4-119B-2603-NVFP4 and Mistral-Small-4-119B-2603-eagle, each ungated and tagged license:apache-2.0.

• https://huggingface.co/api/models/mistralai/Mistral-Small-4-119B-2603 recorded 2026-10-13

  Hub record: "private": false, "gated": false, cardData license "apache-2.0"; files include consolidated-00001-of-00007.safetensors to consolidated-00007-of-00007.safetensors and model-00001-of-00003.safetensors to model-00003-of-00003.safetensors. No dataset tag, and no LICENSE file in the repository.

• https://huggingface.co/mistralai/Mistral-Small-4-119B-2603/raw/main/README.md recorded 2026-10-13

  Card front matter "license: apache-2.0"; "Apache 2.0 License: Open-source license for both commercial and non-commercial use"; "This model is licensed under the Apache 2.0 License". Code covers inference through vLLM, llama.cpp, SGLang and transformers, and fine-tuning points to Axolotl. Searched the body for dataset, training data and corpus: none.

Adoption is measured by Hugging Face downloads of the main checkpoint Mistral publishes. Mistral also serves the model through its API, which publishes no usage figure, so the count covers the open weights alone.

• https://huggingface.co/api/models/mistralai/Mistral-Small-4-119B-2603 recorded 2026-10-13

  "downloads":77411 over the trailing 30 days for mistralai/Mistral-Small-4-119B-2603.

With reasoning on, Mistral Small 4 sits a little below Mistral Medium 3.5 on the Artificial Analysis index and above GPT-4o, far from the frontier but strong for its active size, so it scores level with Mistral Medium.

• https://artificialanalysis.ai/leaderboards/models recorded 2026-10-13

  The leaderboard's embedded model records give intelligenceIndex 14.19 for Mistral Medium 3.5, 12.83 for Nemotron 3 Super 120B A12B (Reasoning), 11.27 for Mistral Small 4 (Reasoning), 8.99 for Mistral Small 4 (Non-reasoning) and 8.44 for GPT-4o (Nov '24). No SEA-LION or Bielik model appears on the page. The digest drifts between fetches because the page carries live pricing and throughput figures.

• https://artificialanalysis.ai/models/mistral-small-4 recorded 2026-10-13

  "Mistral Small 4 (Reasoning) scores 11 on the Artificial Analysis Intelligence Index, placing it well above average among comparable models (median: 8)"; Intelligence rank "# 12 / 65"; tagged "Open weights model" and "Released March 2026".

• https://artificialanalysis.ai/models/mistral-small-4-non-reasoning recorded 2026-10-13

  "Mistral Small 4 (Non-reasoning) scores 9 on the Artificial Analysis Intelligence Index, placing it well above average among comparable models (median: 7)".

• https://docs.mistral.ai/models/mistral-small-4-0-26-03 recorded 2026-10-13

  Mistral's model page: "March 16, 2026", "Apache 2.0", "v 26.03", "Mistral Small 4", "Our powerful hybrid model unifying instruct, reasoning, and coding capabilities in a single model", API identifier "mistral-small-2603", context 256k, priced at $0.15 per million input tokens and $0.6 per million output tokens.