Mistral Small 4 is Mistral AI's open model for fast replies and step-by-step reasoning. It reads images and handles up to 256K tokens in one chat. Mistral Small 4 is available for free on AIFreeForever.
Mistral Small 4 is Mistral AI's open model that gives fast replies or step-by-step reasoning in one model, released on 16 March 2026. It reads text and images, with a 256K-token context window. Chat with Mistral Small 4 free on AIFreeForever.
AIFreeForever is not affiliated with or endorsed by the model providers. All trademarks and logos belong to their owners.
Handles 3× more requests a second than Mistral Small 3.
Scores 71.2 on GPQA Diamond, a test of graduate-level science.
Scores 0.72 on AA-LCR, a test of reasoning over long text.
Reads images as well as text.
and choose Mistral Small 4.
if your question is about a photo, chart or screenshot.
and ask follow-up questions.
Our chatbot gives you Mistral Small 4 free, without a Mistral account.
It handles 3× more requests a second than Mistral Small 3.
Our chatbot switches between quick answers and step-by-step thinking.
When you upload an image, our chatbot can see its content.
Our chatbot takes long reports so you can ask about any part.
Apache 2.0 allows free business use.
AIFreeForever's free chatbot is fully integrated with Mistral Small 4 via the official API providers. To get started, type your question or add a supporting document, and Mistral Small 4 answers in the same chat. Use it to write emails, summarise reports, solve maths questions and more.
Plain chat
Plain chat
Plain chat
Plain chat
@Resume
@Business Mistral Small 4 costs $0.15 per million input tokens and $0.60 per million output tokens on Mistral's API. It's free on AIFreeForever and free to download under Apache 2.0.
Yes. You can use Mistral Small 4 free in three ways:
Mistral Small 4 scored 71.2 on GPQA Diamond and 0.72 on AA-LCR, using about 1.6K characters of output for AA-LCR.
This benchmark was extracted from Mistral AI and Mistral's model card.
Qwen models needed about 5.8K–6.1K characters of output to get a similar AA-LCR score, against about 1.6K for Mistral Small 4. On LiveCodeBench, Mistral Small 4 beat GPT-OSS 120B while writing 20% less.
Mistral Small 4 is the largest Apache 2.0 model below that reads images and switches reasoning on or off.
Mistral Small 4 is a mixture-of-experts model: it has 128 experts and uses 4 for each token, so only 6.5B of its 119B parameters run at a time. Developers set `reasoning_effort` to "none" for fast replies or "high" for step-by-step thinking, as explained in Mistral's model card.
Mistral Small 4 has three limits on AIFreeForever, which connects to it through its API and adds our skills and file editors:
These limitations only cover what you can do with Mistral Small 4 on AIFreeForever, not the limits of the model on Mistral AI’s own platform. All trademarks belong to their owners, and the model’s full features are only available when you use it officially through Mistral AI.
Yes. It's free on AIFreeForever and free to download under Apache 2.0.
Mistral AI made Mistral Small 4.
Mistral AI released Mistral Small 4 on 16 March 2026.
Yes. It reads text and images, and answers in text.
Mistral Small 4 takes up to 256K tokens.
Yes. Developers set it to "none" for fast replies or "high" for step-by-step thinking.
Mistral Small 4 is larger, reads images, adds reasoning and has a 256K-token window. Mistral Small 3 has a 32K-token window and reads text only.
No. It can make mistakes, so check important facts and figures.
Ask Mistral Small 4 a question, paste your code or add a file and get instant answers. Not satisfied? Compare your answers to other models for free.
Chat with Mistral Small 4Tell us if this AI content is offensive, harmful or wrong. We review every report and use it to improve our filters.