Definition
Plain language
An openly available AI system from Mistral that handles both text and pictures.
As stated in the literature
Mistral's open-weight multimodal model family, combining a native vision encoder with the Mistral text backbone for interleaved image-text input.
Why it matters: Open-weight multimodal models like this let anyone run image-and-text experiments on their own hardware, without a vendor's filters sitting in between.
For example, you can hand Pixtral a screenshot of a spreadsheet and a written question about it, and it reads both together.
Heard on the show
“Pixtral is asked to help write a phishing message.”Episode 277 — The Blank White Square That Swings AI Refusal Rates Fifty Points