Llama-3.3-70B-Instruct
unsloth/Llama-3.3-70B-Instruct
70BLlama 33 Heretic artifacts1 measured conversions30.2K downloads
Best version95.2Pareto
0xA50C1A1/Llama-3.3-70B-Instruct-SOM-MPOAby 0xA50C1A1 · Heretic v1.2.0 · GGUF, SAFETENSORS
- KL divergence
- 0.049
- Refusals
- 5/100
- Refusal rate
- 5%
- Downloads
- 30.2K
Evidence from the model card
| KL divergence | 0.0494 | | Refusals | 5/100 |
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.
| # | Author | Params | KL | Refusals | Score | Evidence | Downloads |
|---|---|---|---|---|---|---|---|
| 1 | 0xA50C1A1 | 70B | 0.049 | 5/1005% | 95.2 | BestParetoGGUFSAFETENSORS | 30.2K |
All artifacts (3)
| Repository | Type | Formats | Metrics | Downloads | Updated |
|---|---|---|---|---|---|
| Llama-3.3-70B-Instruct-SOM-MPOA-i1-GGUF mradermacher | quant | GGUFIQ1_M | inheritedKL 0.049 | 24.7K | 2026-04-23 |
| Llama-3.3-70B-Instruct-SOM-MPOA 0xA50C1A1 | heretic | SAFETENSORS | published (weak match)KL 0.049 | 3.9K | 2026-04-20 |
| Llama-3.3-70B-Instruct-SOM-MPOA-GGUF mradermacher | quant | GGUFIQ4_XS | inheritedKL 0.049 | 1.6K | 2026-04-22 |