Ministral-3-8B-Instruct-2512 is an 8-billion parameter instruction-tuned language model released on the Hugging Face platform by Mistral AI. It serves as the underlying model for the Le Chat AI assistant and functions as a foundation model for conditional generation tasks.
The model uses the Mistral3ForConditionalGeneration architecture with a mistral3 model type. Its configuration includes FP8 quantization that excludes conversion for the vision tower, multi-modal projector, and language modeling head. The tokenizer defines specific tokens for beginning-of-sequence, end-of-sequence, padding, and unknown. A custom Jinja chat template is provided that sets a default system message identifying the model as Ministral-3-8B-Instruct-2512, an LLM created by the French startup Mistral AI headquartered in Paris, which powers Le Chat and whose knowledge cutoff is October 2023.
It is hosted as a repository on Hugging Face, allowing download and use through the platform's ecosystem for inference or further development. The page includes standard Hugging Face tags such as mistral-common and indicates the presence of a model card.
Ministral 3 8B Instruct 2512 is a Foundation models & chat product. It focuses on providing a small, efficient, and capable instruction-following model for on-device or local deployment. Ministral 3 8B Instruct 2512 is an open-source project aimed at developers and enterprises. The project is open source (Apache-2.0). Ministral 3 8B Instruct 2512 is available on the web, the command line, and API.
Behind Ministral 3 8B Instruct 2512 is Mistral AI, based in France, and the product first shipped in 2023. The project is developed in the open on GitHub with 86.8k stars and 3k commits in the last 90 days. Among its 3 catalogued features are Instruction Tuning, Multimodal Support, and Le Chat Integration.
Latest indexed changes and source events
mistralai/Ministral-3-8B-Instruct-2512 verified by the PulseGate indexer
Other apps tracked under the same category.
