Rytability Scan Text - Qwen 3.5 4B vision tower

The vision tower of Qwen/Qwen3.5-4B, separated from the text weights so the Rytability app can ship the language model on its own and download vision only when the Scan Text ability is installed.

Source: mlx-community/Qwen3.5-4B-4bit (MLX format, sha256 5fb9acd0246866381cf8c5c354c6db1019f6498eec4ccb4f5edcc71ffeacb2db)
Contents: every vision_tower.* tensor (297 tensors, 333.5M parameters), bfloat16, unquantized, MLX safetensors metadata {"format": "mlx"}
File: model-vision.safetensors (667,061,465 bytes)

At load time the app places this file next to the bundled model-text.safetensors and loads both with mlx-swift-lm's Qwen35 VLM. Without it, the text model loads alone through MLXLLM.

License

Qwen 3.5 is released by the Qwen team at Alibaba Cloud under the Apache License, Version 2.0. These weights are unmodified apart from being separated from the text weights (and converted to MLX format by mlx-community). See https://www.apache.org/licenses/LICENSE-2.0.

Downloads last month: -; Downloads are not tracked for this model. How to track

MLX

Hardware compatibility

Quantized

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Rybib/rytability-qwen35-vision

Base model

Qwen/Qwen3.5-4B-Base

Finetuned

Qwen/Qwen3.5-4B

Finetuned

(865)

this model