Moonlight-16B-A3B-Instruct
moonshotai/Moonlight-16B-A3B-Instruct
16B · 3B activeMoonlight3 Heretic artifacts1 measured conversions22.3K downloads
- KL divergence
- 0.330
- Refusals
- 76/104
- Refusal rate
- 73%
- Downloads
- 22.3K
Evidence from the model card
| **Refusal Reduction** | 76/104 prompts answered (73% success rate) | … | **KL Divergence** | 0.33 (low divergence = capabilities preserved) |
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 | rawcell | 16B · 3B active | 0.330 | 76/10473% | 51.1 | BestParetoGGUFSAFETENSORS | 22.3K |
All artifacts (3)
| Repository | Type | Formats | Metrics | Downloads | Updated |
|---|---|---|---|---|---|
| Moonlight-16B-A3B-Instruct-bruno-i1-GGUF mradermacher | quant | GGUFIQ1_M | inheritedKL 0.330 | 20K | 2026-02-06 |
| Moonlight-16B-A3B-Instruct-bruno-GGUF mradermacher | quant | GGUFIQ4_XS | inheritedKL 0.330 | 2K | 2026-02-06 |
| Moonlight-16B-A3B-Instruct-bruno rawcell | heretic | SAFETENSORS | published (weak match)KL 0.330 | 320 | 2026-02-05 |