Ministral 3 8B instruct model published in BF16 precision
Mistral AI published the BF16-precision variant of the larger Ministral-3-8B instruction-tuned model on Hugging Face under Apache-2.0. The 8B variant offers more capacity than the 3B sibling while sharing the same mistral3 architecture and multi-language support.
PUBLISHED2025-10-31
OBSERVED2026-08-11
AGE9mo
SOURCES1
- Model architecture: mistral3
- Precision: BF16 (safetensors)
- License: Apache-2.0
- Languages: en, fr, es, de, it, pt, nl, zh, ja, ko, ar
- Paper: arxiv:2601.08584
- Base model: mistralai/Ministral-3-8B-Base-2512 (finetuned)
COMMUNITY
No curated reactions recorded for this event. Facts and takes are kept in separate layers — community context is added by hand, never blended into the record above.