Ministral 3 14B Instruct 2512
Mistral AI · 2025-12-02 · 13.9B parameters
Ministral 3 14B Instruct 2512 is Mistral AI's FP8 instruction-post-trained dense multimodal checkpoint, released in December 2025 with a 13.5B language model and a 410M-parameter vision encoder. It accepts text and image input, supports a 262,144-token context via YaRN extension from 16,384 positions, and provides multilingual chat, native function calling, and JSON output; it is distinct from the separately released Base and Reasoning variants. Mistral's API serves it as ministral-14b-2512 (model card: https://docs.mistral.ai/models/model-cards/ministral-3-14b-25-12).