Minerva-7B-Instruct-v1.0-MLX-4bit

MLX 4-bit quantisation of sapienzanlp/Minerva-7B-instruct-v1.0, converted for use on Apple Silicon via mlx-lm.

Source model

• Repository: sapienzanlp/Minerva-7B-instruct-v1.0
• Release: 2024-11
• Family: minerva
• Origin: eu
• Languages / coverage: Italian + English + code, ~50/50 IT/EN training mix
• License: apache-2.0 (inherited)

Notes from upstream

Sapienza NLP, FAIR, CINECA. Italy's first LLM family pretrained from scratch on Italian. ~2.5T tokens (1.14T IT, 1.14T EN, 200B code).

Conversion details

• Tool: mlx-lm 0.31.3
• Quantisation: 4-bit (defaults from mlx_lm.convert)
• Converted on: 2026-05-05

Usage

from mlx_lm import load, generate

model, tokenizer = load("luiscalisto/Minerva-7B-Instruct-v1.0-MLX-4bit")
prompt = "Hello, who are you?"
print(generate(model, tokenizer, prompt=prompt, max_tokens=128, verbose=False))

License and attribution

This is a quantised redistribution of sapienzanlp/Minerva-7B-instruct-v1.0. The original model and its license terms (apache-2.0) carry through unchanged. Please cite the upstream authors when using this model. See the source repository for the authoritative model card and citation.

Conversion provenance

Produced by llm-mlx-conversions, a small utility for publishing community MLX 4-bit quants of open-weight LLMs.