API Try Docs Blog

LFM3-VL-2B-MLX-4bit

4-bit MLX version of LFM3-VL-2B for Apple Silicon. The weights take 1.59 GB.

Weights use 4-bit affine quantization with groups of 32. Token embeddings stay in 8-bit. The weights are trained with quantization-aware distillation (QAD), so they recover much of the 8-bit performance.

Find architecture and usage details in the original model card.

Use with MLX-VLM

---

Requires MLX-VLM 0.7.11 or later:

pip install -U "mlx-vlm>=0.7.11"

python -m mlx_vlm.generate --model LiquidAI/LFM3-VL-2B-MLX-4bit \
  --image <IMAGE PATH> --prompt "Describe this image." --enable-thinking \
  --temperature 0.1 --top-p 1.0 --top-k 50 --min-p 0.0 --repetition-penalty 1.1 \
  --max-tokens 16384

Reasoning stays off unless you pass --enable-thinking. More options in the MLX guide.