ThinkingCap-Qwen3.6-27B
bottlecapai/ThinkingCap-Qwen3.6-27B
27BQwen 3.63 Heretic artifacts1 measured conversions10.1K downloads
Best version93.1Pareto
egibson/ThinkingCap-Qwen3.6-27B-hereticby egibson · Heretic v1.4.0 · GGUF, SAFETENSORS
- KL divergence
- 0.019
- Refusals
- 12/100
- Refusal rate
- 12%
- Downloads
- 9.1K
Evidence from the model card
| | Keywords (refusals detected on harmful prompts) | KL divergence from original model on harmless prompts | → | This model | 12/100 | 0.0186 |
Metrics published by the author. Why this ranks first.
All measured versions
Higher score is better. One row per ablation attempt; artifacts of the same attempt are grouped under it.
| # | Author | Params | KL | Refusals | Score | Evidence | Downloads |
|---|---|---|---|---|---|---|---|
| 1 | egibson | 27B | 0.019 | 12/10012% | 93.1 | BestParetoGGUFSAFETENSORS | 9.1K |
All artifacts (3)
| Repository | Type | Formats | Metrics | Downloads | Updated |
|---|---|---|---|---|---|
| ThinkingCap-Qwen3.6-27B-heretic-GGUF egibson | quant | GGUFIQ2_M | inheritedKL 0.019 | 9K | 2026-07-20 |
| ThinkingCap-Qwen3.6-27B-abliterated-NVFP4 maci0 | heretic | GPTQNVFP4SAFETENSORSNVFP4 | noneKL 0.041 | 1K | 2026-08-04 |
| ThinkingCap-Qwen3.6-27B-heretic egibson | heretic | SAFETENSORS | published (weak match)KL 0.019 | 104 | 2026-07-19 |
1 artifact inherit metrics from their parent conversion or publish none.