AI & Machine Learning / Image Generation Verified 26 Apr 2026

Mistral Le Pixtral

Multimodal image+text from Mistral

Paris, France · Founded 2023 · Freemium
A European alternative to Midjourney Runway Leonardo.AI Stability AI Adobe Firefly Ideogram

About Mistral Le Pixtral

Pixtral 12B is the first open-source multimodal model from Mistral AI, the French AI lab headquartered in Paris. Released in September 2024 under Apache 2.0, it pairs a 12B-parameter language backbone with a 400M-parameter vision encoder, supports a 128K-token context, and ingests any number of images at native resolution and aspect ratio. It targets image understanding, OCR, chart and document analysis and visual question answering while preserving Mistral's strong text, instruction-following, coding and math behaviour. Weights are free to download and self-host; the Mistral API charges roughly USD 0.15 per million input and output tokens.

Quick facts

Pricing model Freemium
Free trial
Open source
Self-hostable No
Public API No
Target audience B2B
Company size fit All
DPA available Yes
Data stored in EU Yes
Last verified by eubuilt.eu on 26 Apr 2026 — re-checks every 6 months Flag listing → Read our criteria →