Llama3.1-SuperHawk-8B
Yuma42/Llama3.1-SuperHawk-8B
8BLlama 3.16 Heretic artifacts2 measured conversions19.9K downloads
Best version95.2Pareto
ChiKoi7/Llama3.1-SuperHawk-8B-Heretic-v2by ChiKoi7 · Heretic v1.1.0 · GGUF, SAFETENSORS
- KL divergence
- 0.049
- Refusals
- 5/100
- Refusal rate
- 5%
- Downloads
- 9.4K
Evidence from the model card
| Refusals | 5/100 | | KL divergence | 0.0493 |
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.
All artifacts (6)
| Repository | Type | Formats | Metrics | Downloads | Updated |
|---|---|---|---|---|---|
| Llama3.1-SuperHawk-8B-Heretic-i1-GGUF mradermacher | quant | GGUFIQ1_M | none | 7.6K | 2025-12-24 |
| Llama3.1-SuperHawk-8B-Heretic-v2-i1-GGUF mradermacher | quant | GGUFIQ1_M | inheritedKL 0.049 | 5.3K | 2025-12-24 |
| Llama3.1-SuperHawk-8B-Heretic-v2-GGUF mradermacher | quant | GGUFIQ4_XS | inheritedKL 0.049 | 3.6K | 2025-12-24 |
| Llama3.1-SuperHawk-8B-Heretic-GGUF mradermacher | quant | GGUFIQ4_XS | none | 2.8K | 2025-12-24 |
| Llama3.1-SuperHawk-8B-Heretic-v2 ChiKoi7 | heretic | SAFETENSORS | published (weak match)KL 0.049 | 491 | 2025-12-23 |
| Llama3.1-SuperHawk-8B-Heretic ChiKoi7 | merge | SAFETENSORS | published (weak match)KL 0.095 | 69 | 2025-12-23 |
2 artifacts inherit metrics from their parent conversion or publish none.