Qwen3.8-27B-Humanlike-Chat-2.0 EXL3 4.00 bpw

Source: https://huggingface.co/LessThanThreeAI/Qwen3.8-27B-Humanlike-Chat-2.0

Source revision: 0a4d369895bf0a8b217fb08ef738769bac9d881b

ExLlamaV3 revision: 1.5.2+cu128.torch2.10.0

Target bitrate: 4.00 bpw

Architecture: Qwen3_5ForConditionalGeneration

This conversion was produced with ExLlamaV3's official convert.py and passed a local load/generation smoke test before publication.

Local validation

• Generated tokens: 25
• Decode speed: 35.287 tok/s
• GPU: NVIDIA GeForce RTX 4090
• Torch: 2.10.0+cu128

The speed figure is a local smoke benchmark, not a cross-system performance claim.