Bielik-4.5B-v3.0-Instruct-MLX-4bit

MLX format conversion of speakleash/Bielik-4.5B-v3.0-Instruct for Apple Silicon with 4-bit quantization.

Model Description

Bielik-4.5B-v3.0-Instruct is a Polish instruction-tuned language model based on the Qwen architecture, trained by SpeakLeash. This is the 4-bit quantized MLX conversion optimized for Apple Silicon Macs with reduced memory footprint.

Usage

from mlx_lm import load, generate

model, tokenizer = load("czlonkowski/Bielik-4.5B-v3.0-Instruct-MLX-4bit")
response = generate(model, tokenizer, prompt="Wyjaśnij czym jest uczenie maszynowe.", max_tokens=200)
print(response)

Or via CLI:

mlx_lm.generate --model czlonkowski/Bielik-4.5B-v3.0-Instruct-MLX-4bit --prompt "Wyjaśnij czym jest uczenie maszynowe."

Model Details

• Base Model: speakleash/Bielik-4.5B-v3.0-Instruct
• Format: MLX (Apple Silicon optimized)
• Quantization: 4-bit (~4.5 bits per weight)
• Parameters: 4.5B
• Type: Instruction-tuned

Related Models

• czlonkowski/Bielik-4.5B-v3.0-Instruct-MLX - Full precision (bfloat16) version

License

Please refer to the original model's license at speakleash/Bielik-4.5B-v3.0-Instruct.