Qwen3.8-27B-EXL3 2.0bpw (Mirror)

⚠️ This is a mirror repository. All credit for the quantization goes to turboderp.

Original repository: turboderp/Qwen3.8-27B-exl3 (branch: SC_2.00bpw_H3_V3)

This repo exists solely to provide an alternative download source. I did not create this quant — it is an exact copy of turboderp's work.

---

Model Details

• Base model: Qwen/Qwen3.8-27B
• Quantization: EXL3, 2.0 bits per weight (SC_2.00bpw_H3_V3)
• Quantized by: turboderp
• Format: EXL3 (ExLlamaV3)
• Size: ~9.7 GB

About

This is an EXL3 quantization of Qwen3.8-27B, a native vision-language model with 27B parameters. The quantization was performed by turboderp using the ExLlamaV3 framework.

For usage instructions, benchmarks, and technical details, please refer to the original repository.

Usage

Use with ExLlamaV3 or compatible inference engines (e.g., text-generation-webui with EXL3 support).

License

Apache-2.0 (inherited from the base model)