H2O-Lightning-4B — MLX 4-bit

A 4-bit MLX conversion (affine, group size 64; mlx-vlm 0.7.6) of h2oai/h2o-lightning-4b v1.2.3, for Apple Silicon. 3.0 GB.

Published by NetProjects (Internet Projects Ltd). Not affiliated with H2O.ai.

This is a decision model, not a chat model. It returns a probability for each answer option, not text.

Serve it with the H2O shim, not with mlx_lm.server or a plain mlx_vlm generate call. The generic "Use this model" snippets Hugging Face shows for MLX models do not return the label probabilities this model is made for. The commands below set it up with H2O's official shim, unmodified, on top of h2o-lightning-mlx, which reads the label log-probabilities with a float32 head, as the original does on vLLM.

The higher score is noise (11 answers flip both ways) and calibration is visibly looser. Prefer the 8-bit build unless memory is tight.

Images. Image input works through the unchanged shim. On H2O's own receipt example this build gives the same answers as the model card shows. In a small generated-image check (counting, colours, reading a table) it was slightly weaker at counting objects (28 of 30 against 30 of 30 for the 8-bit build). It is not a reproduction of H2O's image benchmark.

Limits (as for the original): mostly evaluated in English; works best with about 16 answer options or fewer; no explanations and no video input.

git clone https://github.com/internet-projects/h2o-lightning-mlx && cd h2o-lightning-mlx && ./setup.sh
hf download netprojects/h2o-lightning-4b-mlx-q4 --local-dir mlx-q4
MODEL=mlx-q4 ./serve.sh # http://127.0.0.1:8741/v1/systemone

Tested with the shim at H2O revision v1.2.3.

Full precision: use the original weights, h2oai/h2o-lightning-4b, with the same repo and shim. There is no separate bf16 copy here.

Credits: H2O.ai (H2O-Lightning-4B, Apache-2.0), Qwen team (Qwen3.5-4B, Apache-2.0). This conversion is not affiliated with H2O.ai.