Other

Mistral releases open-source Small 4 model with hybrid reasoning and image understanding

Original title:Mistral has released Mistral Small 4, an open weights model with hybrid reasoning and image input, s...

AI summary

Mistral released the open-weight Mistral Small 4, a 119B-parameter MoE architecture (6.5B active parameters per token) supporting switchable reasoning/non-reasoning modes and image input. In reasoning mode it scored 27 on the Artificial Analysis Intelligence Index, beating Mistral Large 3 but trailing competitors such as gpt-oss-120B. The model is more token-efficient than peers with a lower hallucination rate (AA-Omniscience -30), supports a 256K context window, and is released under the Apache 2.0 license.

Key Takeaways

  • Mistral released the open-weight Mistral Small 4,a 119B-parameter MoE architecture (6.
  • 5B active parameters per token) supporting switchable reasoning/non-reasoning modes and image input.
  • In reasoning mode it scored 27 on the Artificial Analysis Intelligence Index,beating Mistral Large…
  • The model is more token-efficient than peers with a lower hallucination rate (AA-Omniscience…
  • 0 license.

Quick Facts

Published
Updated
Source
X:Artificial Analysis (@ArtificialAnlys)
Category
Other
Read the original Read the full article at the original source
Why it matters

Mistral 开源 Small 4,支持混合推理与多模态,Agent 任务表现大幅提升

Related entities

Source & copyright

This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.

Original article · AIHOT entry

How to Cite

When citing this article, keep the title, date, and permalink, and credit Orange News as the source.

Orange News: Mistral releases open-source Small 4 model with hybrid reasoning and image understanding. March 20, 2026. https://new.nat6.net/en/news/cmnw1ypby00oislc3fue2oy4c/

View all →