| --- |
| license: apache-2.0 |
| language: |
| - en |
| - es |
| - zh |
| tags: |
| - mlx |
| - vlm |
| - vision |
| - osirisbrain |
| - apple-silicon |
| - qwen3.5 |
| base_model: Qwen/Qwen3.5-9B |
| pipeline_tag: image-text-to-text |
| library_name: mlx |
| --- |
| |
| # OsirisCortex-v7-Censo-MLX |
|
|
| **Cortex v7 (Censored + Vision)** — Osiris's main reasoning + vision brain. VLM (Vision-Language Model) that can process images, screenshots, and video frames. Runs natively on Apple Silicon via MLX Metal. |
|
|
| ## Architecture |
|
|
| - **Base Model:** Qwen3.5-9B VLM (9B params, vision + language) |
| - **Architecture:** `Qwen3_5ForConditionalGeneration` (multimodal) |
| - **Format:** MLX mxfp4 quantized (Apple Silicon native) |
| - **Size:** ~5.3 GB (includes vision encoder) |
| - **Vision:** Full image understanding, OCR, screenshot analysis, video frames |
| - **Note:** This is the censored (base) version. For uncensored, see OsirisCortex-v7-MLX. |
|
|
| ## Usage |
|
|
| ```python |
| from mlx_vlm import load, generate |
| |
| model, processor = load("osirisbrain/OsirisCortex-v7-Censo-MLX") |
| output = generate(model, processor, "Describe this image", ["path/to/image.jpg"]) |
| ``` |
|
|
| ## Credits |
|
|
| Converted by [RepublicOfKorokke](https://huggingface.co/RepublicOfKorokke/Qwen3.5-9B-mlx-vlm-mxfp4). |
| Original model: [Qwen/Qwen3.5-9B](https://huggingface.co/Qwen/Qwen3.5-9B) by Alibaba. |
|
|
