release aidev NEEDS REVIEW SIG 1/5

Ministral 3 8B instruct model published in FP8 quantized format

Mistral AI published the FP8 quantized variant of the Ministral-3-8B instruction-tuned model on Hugging Face under Apache-2.0. A quantized version of the 8B base instruct model, suitable for deployment with reduced memory footprint.

PUBLISHED2025-10-31
OBSERVED2026-08-11
AGE9mo
SOURCES1
  • Model architecture: mistral3
  • Precision: FP8 (quantized)
  • License: Apache-2.0
  • Languages: en, fr, es, de, it, pt, nl, zh, ja, ko, ar
  • Base model: mistralai/Ministral-3-8B-Base-2512 (quantized)
model-support

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.