Hugging Face's logo

Join the conversation

Join the community of Machine Learners and AI enthusiasts.

Sign Up
Back to feed
DavidAU 
posted an update 27 days ago
Post
22419
Qwen3.8 - 27B - COLD FUSION

Reduction in thinking tokens to 1/10 to 1/2 "normal Qwen" across all three modes of thinking without loss of detail in thinking or output.

Increase in general intelligence too (all 7 critical benches); bench marks posted.

Trained using the COLD FUSION method (GAIN + UNSLOTH).
NOTE: This is NOT a heretic / uncensored version.

MTP and GGUF NEO MAX quants:
DavidAU/Qwen3.8-27B-Cold-Fusion-GAIN-V1.1-NM-DAU-NEO-MAX-MTP-GGUF

SOURCE:
DavidAU/Qwen3.8-27B-Cold-Fusion-GAIN-V1.1
gameonvinay

The Source files are 3 days old I hope they are updated. Thank you!
Anonymuz
This comment has been hidden (marked as Graphic Content)
Anonymuz

⭐⭐⭐⭐⭐
Nevermore9898

I've been running some independent benchmarking on Cold Fusion Q4 MTP using lm-evaluation-harness. Early results on the generative pass:

IFEval prompt-strict: 86.5% (loose: 88.7%)
GSM8K flexible: 92.9%

MATH-hard got zeroed out due to a harness issue — the Problem: stop string was killing the thinking trace before the model could output an answer. Not a model problem, confirmed by inspecting the captures. Planning a clean rerun with adjusted stop strings and a higher token cap.

Still learning the eval tooling, but more numbers coming once I finish the other Q4 variants and give CF a fair rerun on the affected tasks!
Exnsr

Will you release more 16gb-friendly quantizations in the future on this training? Amazing job you've done by the way!

·
DavidAU

You can find additional quants here:
https://huggingface.co/models?other=base_model:quantized:DavidAU/Qwen3.8-27B-Cold-Fusion-GAIN-V1.1

Some are a lot smaller / will work better on 16GB cards.
privacydied

cant wait for the uncensored!!!
In this post