osmapi
/

Step-3.7-Flash-MXFP4-mlx

Image-Text-to-Text

vision-language

Mixture of Experts

text-generation

4-bit precision

Model card Files Files and versions

junafinity commited on about 21 hours ago

Commit

a4eb4e0

·

verified ·

1 Parent(s): 42b40a9

Add public Step-3.7 variant links

Files changed (1) hide show

README.md +7 -0

README.md CHANGED Viewed

@@ -30,6 +30,13 @@ tags:
 No fine-tuning, distillation, or retraining was applied. The upstream StepFun checkpoint was downloaded and verified locally, then eligible text and vision `.weight` tensors were converted with MLX `mode="mxfp4"` quantization. Tokenizer, chat template, custom Step3.7 Python modules, and non-quantized control tensors are preserved from the source release.
 ## Compatibility Status
 This upload is a standard MLX MXFP4 safetensors bundle, but it is not yet a drop-in `mlx_lm.load(...)` or `mlx_vlm.load(...)` model.

 No fine-tuning, distillation, or retraining was applied. The upstream StepFun checkpoint was downloaded and verified locally, then eligible text and vision `.weight` tensors were converted with MLX `mode="mxfp4"` quantization. Tokenizer, chat template, custom Step3.7 Python modules, and non-quantized control tensors are preserved from the source release.
+## Public osmAPI Step-3.7-Flash Variants
+| Variant | Repository | Format | Notes |
+|---|---|---|---|
+| MXFP4 MLX | [osmapi/Step-3.7-Flash-MXFP4-mlx](https://huggingface.co/osmapi/Step-3.7-Flash-MXFP4-mlx) | MLX MXFP4 safetensors | Public 4-bit microscaling tensor bundle |
+| OptiQ 3.7bpw MLX | [osmapi/Step-3.7-Flash-OptiQ-3.7bpw-mlx](https://huggingface.co/osmapi/Step-3.7-Flash-OptiQ-3.7bpw-mlx) | MLX affine mixed-precision safetensors | Public 3.7 BPW OptiQ tensor bundle |
 ## Compatibility Status
 This upload is a standard MLX MXFP4 safetensors bundle, but it is not yet a drop-in `mlx_lm.load(...)` or `mlx_vlm.load(...)` model.