Qwen3-Coder-Next
Qwen/Qwen3-Coder-Next
79.7BQwen 39 Heretic artifacts2 measured conversions94.3K downloads
Best version89.7Pareto
trohrbaugh/Qwen3-Coder-Next-hereticby trohrbaugh · Heretic v1.2.0 · GGUF, SAFETENSORS
- KL divergence
- 0.071
- Refusals
- 14/100
- Refusal rate
- 14%
- Downloads
- 59.4K
Evidence from the model card
| KL divergence | 0.0708 | | Refusals | 14/100 |
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 | trohrbaugh | 79.7B | 0.071 | 14/10014% | 89.7 | BestParetoGGUFSAFETENSORS | 59.4K |
| 2 | llmfan46 | 79.7B | 0.149 | 8/1008% | 89.5 | ParetoGGUFSAFETENSORS | 9.2K |
All artifacts (9)
| Repository | Type | Formats | Metrics | Downloads | Updated |
|---|---|---|---|---|---|
| Qwen3-Coder-Next-heretic-i1-GGUF mradermacher | quant | GGUFIQ1_M | inheritedKL 0.071 | 42.5K | 2026-03-05 |
| Qwen3-Coder-Next-Uncensored-Heretic-i1-GGUF mradermacher | quant | GGUFIQ1_M | none | 21.2K | 2026-08-31 |
| Qwen3-Coder-Next-heretic trohrbaugh | heretic | SAFETENSORS | published (weak match)KL 0.071 | 9.4K | 2026-04-14 |
| Qwen3-Coder-Next-Uncensored-Heretic-GGUF llmfan46 | quant | GGUFBF16 | published (weak match)KL 0.149 | 8.5K | 2026-08-30 |
| Qwen3-Coder-Next-heretic-GGUF mradermacher | quant | GGUFIQ4_XS | inheritedKL 0.071 | 4.8K | 2026-03-04 |
| Qwen3-Coder-Next-Uncensored-Heretic-GGUF mradermacher | quant | GGUFIQ4_XS | none | 4.5K | 2026-08-30 |
| Qwen3-Coder-Next-heretic-nvfp4 trohrbaugh | heretic | NVFP4SAFETENSORSNVFP4 | published (weak match)KL 0.071 | 1.7K | 2026-04-14 |
| Qwen3-Coder-Next-Heretic-GGUF ghecko78 | quant | GGUFQ4_K_M | inheritedKL 0.071 | 1K | 2026-03-03 |
| Qwen3-Coder-Next-Uncensored-Heretic llmfan46 | finetune | SAFETENSORS | published (weak match)KL 0.149 | 744 | 2026-08-30 |
2 artifacts inherit metrics from their parent conversion or publish none.