gemma-4-E4B-it-heretic-mlx-nvfp4
Converted from coder3101/gemma-4-E4B-it-heretic with mlx_vlm.convert.
- Quantization: nvfp4
- q_bits: 4
- q_mode: nvfp4
- Source type: Gemma 4 multimodal (Gemma4ForConditionalGeneration)
- Converter env: /Users/vanch/.cache/lm-studio/.venvs/gemma4-mlx
- Validation date: 2026-04-26
Validation Summary
- Status: passed local text and vision smoke tests
- Vision metadata present:
processor_config.jsonandpreprocessor_config.json - Peak memory across sampled tests: 7.669 GB
Validation Matrix
| Case | Output | Prompt TPS | Generation TPS | Peak Memory |
|---|---|---|---|---|
| Text: one-sentence model description | Gemma 4 is a family of open-weights large language models developed by Google DeepMind. | 173.435 | 67.026 | 6.921 GB |
| Text: two-bullet structured answer | * This model belongs to the Gemma family of open-weights models. * As a member of this family, it can process and understand multimodal inputs, depending on the specific version being utilized. |
238.298 | 65.270 | 6.933 GB |
| Vision: red image color identification | The image is red. | 576.505 | 73.123 | 7.627 GB |
| Vision: blue image color identification | The image is blue. | 570.144 | 73.028 | 7.669 GB |
Local Artifacts
- Full verification logs:
/Users/vanch/models/gemma4-e4b-heretic-full-verify/nvfp4-*.log