DavidAU/Qwen3.8-27B-TWIN-TURBO-Fable-Cold-Fusion-709-L-Uncensored-NM-DAU-NEO-MTP-GGUF — new model trending #30 on Hugging Face
DavidAU releases Qwen3.8-27B-TWIN-TURBO, an uncensored GGUF fine-tune claiming sharply reduced thinking tokens and self-reported benchmark gains.
A new community fine-tune of Qwen3.8-27B, built with Unsloth via 'Cold Fusion' and 'Fable Fusion 711' methods, is trending #30 on Hugging Face. The author claims arc-c scores of 709 in 8-bit and 701 in 4-bit, exceeding the base Qwen3.8-27B, while cutting thinking tokens to between 1/2 and 1/20 of baseline. The release provides regular and MTP Neo/NEO MAX GGUF quants, five switchable reasoning modes plus five zero-reasoning instruct modes, and a lightly uncensored variant. All benchmark claims are author-reported and not independently verified.
- 27B multi-stage fine-tune/merge of Qwen3.8 distributed as GGUF quants.
- Claims arc-c 709 (8-bit) and 701 (4-bit), above base Qwen3.8-27B.
- Reduces thinking tokens by 1/2 to as low as 1/20 versus base model.
- Five reasoning and five instruct modes switchable via API or in chat.
- Lightly uncensored; benchmark numbers are self-reported.
- Vendors
- DavidAU
- Organizations
- Hugging Face
From the original article
14 min read at huggingface.coimage-text-to-text · 133,885 downloads · 157 likes · license apache-2. 0
By DavidAU · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.