exolabs/qwen3-6-35b-a3b-ud-mlx-4bit-text-dequant-bf16
The exolabs/qwen3-6-35b-a3b-ud-mlx-4bit-text-dequant-bf16 is a 35.1 billion parameter model derived from the Qwen3.6-35B-A3B architecture. This specific checkpoint is a dequantized bfloat16 export of an MLX 4-bit quantized model, not the original upstream bfloat16 version. It is intended for vLLM validation, focusing on its specific dequantized bfloat16 format.
Loading preview...
Model Overview
This model, exolabs/qwen3-6-35b-a3b-ud-mlx-4bit-text-dequant-bf16, is a 35.1 billion parameter variant based on the Qwen/Qwen3.6-35B-A3B architecture. It represents a specific export format, distinct from the original upstream model.
Key Characteristics
- Parameter Count: 35.1 billion parameters.
- Source: Derived from
unsloth/Qwen3.6-35B-A3B-UD-MLX-4bit, which is an MLX 4-bit quantized version. - Export Type: This checkpoint is a dequantized bfloat16 (BF16) export of the aforementioned MLX quantized model.
- Distinction: It is crucial to note that this is not the original BF16 checkpoint but rather a dequantized version of a previously quantized MLX model.
- Context Length: Supports a context length of 32768 tokens.
Primary Use Case
This specific model checkpoint is primarily intended for vLLM validation, focusing on evaluating its performance and characteristics in its dequantized bfloat16 format within the vLLM framework.