DeepSeek-R1-Distill-Llama-70B
deepseek-ai/DeepSeek-R1-Distill-Llama-70B
70BLlama5 Heretic artifacts2 measured conversions165.1K downloads
Best version90.4Pareto
Total04/DeepSeek-R1-Distill-Llama-70B-hereticby Total04 · Heretic v1.2.0 · GGUF, SAFETENSORS
- KL divergence
- 0.033
- Refusals
- 16/100
- Refusal rate
- 16%
- Downloads
- 163.1K
Evidence from the model card
| KL divergence | 0.0333 | | Refusals | 16/100 |
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 (5)
| Repository | Type | Formats | Metrics | Downloads | Updated |
|---|---|---|---|---|---|
| DeepSeek-R1-Distill-Llama-70B-heretic-GGUF mradermacher | quant | GGUFIQ4_XS | inheritedKL 0.033 | 82.6K | 2026-02-24 |
| DeepSeek-R1-Distill-Llama-70B-heretic-i1-GGUF mradermacher | quant | GGUFIQ1_M | inheritedKL 0.033 | 79.5K | 2026-02-24 |
| DeepSeek-R1-Distill-Llama-70B-heretic ThalisAI | heretic | GGUFSAFETENSORSQ3_K_M | published (weak match)KL 0.036 | 1.9K | 2026-02-18 |
| DeepSeek-R1-Distill-Llama-70B-heretic-GGUF lactroiii | quant | GGUFQ6_K | inheritedKL 0.033 | 951 | 2026-04-05 |
| DeepSeek-R1-Distill-Llama-70B-heretic Total04 | heretic | SAFETENSORS | published (weak match)KL 0.033 | 152 | 2026-02-22 |