This model was quantised primarily for use with OpenMarker — an open source AI assistant designed to help professors, lecturers, and educators automatically grade and review student work.

OpenMarker uses local quantised models so that:

No student data is sent to external APIs
It runs fully offline and privately
Anyone can self-host it for free

If you're an educator or developer, check out the project and contribute! 👉 https://github.com/theDALEX/openmarker

📦 Quant Details

Quant	Size	Description
Q4_K_M	~4.1GB	Good quality, recommended for most use cases

Quantised using llama.cpp release b4821 on Windows.

🔗 Original Model

This is a quantised version of the original open source model by Mistral AI, released under the Apache 2.0 license.

Original weights: mistralai/Mistral-7B-Instruct-v0.3
Original license: Apache 2.0
Changes made: None — weights are unchanged beyond quantisation

dalexdavis/Mistral-7B-Instruct-v0.3-Q4_K_M-GGUF

Mistral-7B-Instruct-v0.3 GGUF (Q4_K_M)

🎓 About OpenMarker

📦 Quant Details

🔗 Original Model

dalexdavis/
Mistral-7B-Instruct-v0.3-Q4_K_M-GGUF