exolabs/qwen3-6-27b-optiq-4bit-text-dequant-bf16

VISIONConcurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 6, 2026License:otherArchitecture:Transformer0.0K Featherless Exclusive Cold

The exolabs/qwen3-6-27b-optiq-4bit-text-dequant-bf16 is a 27 billion parameter language model, derived from the Qwen3.6-27B architecture. This specific version is a dequantized bfloat16 export of a 4-bit quantized MLX checkpoint, intended for vLLM validation. It offers a balance of performance and efficiency, making it suitable for applications requiring a robust language model with optimized inference characteristics.

Loading preview...

Model Overview

This model, exolabs/qwen3-6-27b-optiq-4bit-text-dequant-bf16, is a 27 billion parameter language model based on the Qwen3.6-27B architecture. It represents a dequantized bfloat16 export of an internally quantized MLX checkpoint (qwen3-6-27b-optiq-4bit), originally sourced from mlx-community/Qwen3.6-27B-OptiQ-4bit.

Key Characteristics

  • Parameter Count: 27 billion parameters, offering significant language understanding and generation capabilities.
  • Dequantized Export: This is a bfloat16 export resulting from the dequantization of a 4-bit quantized MLX checkpoint, rather than the original upstream bfloat16 model.
  • Purpose: Primarily intended for vLLM validation, suggesting its use in environments where efficient inference with large language models is critical.
  • Context Length: Supports a context length of 32768 tokens, enabling processing of extensive inputs and generating coherent long-form content.

Use Cases

  • vLLM Validation: Ideal for testing and validating performance within vLLM inference frameworks.
  • Efficient Inference: Suitable for applications requiring a large language model with optimized memory footprint and faster inference due to its dequantized bfloat16 format.
  • Research and Development: Can be used by researchers and developers exploring the performance characteristics of dequantized models derived from 4-bit quantization.