Qwen 2.5
Best conversion per model
One row per base model. The full list of attempts is on each model page.
| # | Model | Author | Params | KL | Refusals | Score | Evidence | Downloads |
|---|---|---|---|---|---|---|---|---|
| 1 | Qwen2.5-72B-Instruct Qwen | krthkyadtya | 72B | 0.0035 | 0/660% | 99.8 | BestParetoGGUFSAFETENSORS | 13.3K |
| 2 | Qwen2.5-Coder-7B-Instruct Qwen | trohrbaugh | 7B | 0.014 | 2/1002% | 98.3 | BestParetoGGUFSAFETENSORS | 18.6K |
| 3 | Qwen2.5-VL-3B-Instruct Qwen | saidutta69 | 3B | 0.018 | 2/1002% | 98.1 | BestParetoGGUFSAFETENSORS | 726 |
| 4 | Qwen2.5-Coder-3B-Instruct Qwen | saidutta69 | 3B | 0.016 | 3/1003% | 97.7 | BestParetoGGUFSAFETENSORS | 18.4K |
| 5 | Qwen2.5-3B-instruct-medical-finetuned Joker-sxj | ChiKoi7 | 3B | 0.013 | 4/1004% | 97.4 | BestParetoGGUFSAFETENSORS | 7.3K |
| 6 | Qwen2.5-Coder-1.5B-Instruct Qwen | sourcepirate | 1.5B | 0.030 | 3/1003% | 97.0 | BestParetoSAFETENSORS | 461 |
| 7 | Qwen2.5-VL-7B-Instruct Qwen | sh0ck0r | 7B | 0.011 | 5/1005% | 97.0 | BestParetoSAFETENSORS | 121 |
| 8 | Qwen2.5-0.5B-Instruct Qwen | megabytes | 0.5B | 0.040 | 3/1003% | 96.6 | BestParetoGGUFSAFETENSORS | 1.6K |
| 9 | fable-qwen2.5-3b-agentic-merged devxyasir | richardyoung | 3B | 0.050 | 3/1003% | 96.1 | BestParetoMLXSAFETENSORS | 676 |
| 10 | Qwen2.5-1.5B-Instruct Qwen | HarethahMo | 1.5B | 0.054 | 3/1003% | 95.9 | BestParetoSAFETENSORS | 2.5K |
| 11 | Qwen2.5-VL-72B-Instruct Qwen | coder3101 | 72B | 0.016 | 8/1008% | 95.2 | BestParetoFP8GGUFSAFETENSORS | 20K |
| 12 | Qwen2.5-Coder-3B Qwen | davidterrell1919 | 3B | 0.063 | 4/1004% | 95.0 | BestParetoSAFETENSORS | 393 |
| 13 | Qwen2.5-7B-Instruct Qwen | saidutta69 | 7B | 0.076 | 3/1003% | 95.0 | BestParetoGGUFSAFETENSORS | 19.1K |
| 14 | Qwen2.5-3B-Instruct Qwen | HarethahMo | 3B | 0.090 | 2/1002% | 94.8 | BestParetoSAFETENSORS | 58 |
| 15 | Qwen2.5-14B-Instruct Qwen | dec0dedd | 14B | 0.070 | 5/1005% | 94.2 | BestParetoSAFETENSORS | 400 |
| 16 | Qwen2.5-32B-Instruct Qwen | dec0dedd | 32B | 0.081 | 5/1005% | 93.7 | BestParetoGGUFSAFETENSORS | 246 |
| 17 | Qwen2.5-14B Qwen | TitleOS | 14B | 0.048 | 9/1009% | 93.2 | BestParetoGGUFSAFETENSORS | 5.9K |
| 18 | Qwen2.5-3B Qwen | Dotalix | 3B | 0.190 | 2/1002% | 91.0 | BestParetoGGUFSAFETENSORS | 4.8K |
| 19 | Qwen2.5-Coder-0.5B-Instruct Qwen | saidutta69 | 0.5B | 0.125 | 8/1008% | 90.5 | BestParetoGGUFSAFETENSORS | 6.5K |
| 20 | Qwen2.5-14B-Instruct-1M Qwen | richardyoung | 14B | 0.280 | 3/1003% | 87.6 | BestParetoSAFETENSORS | 624 |
| 21 | Qwen2.5-VL-32B-Instruct Qwen | coder3101 | 32B | 0.412 | 14/10014% | 78.4 | BestParetoGGUFSAFETENSORS | 12.5K |
All models
Unmeasured models
heretic_DistilQwen2.5-0.5B-Instruct, heretic_EVA-D-Qwen2.5-1.5B-v0.0, heretic_EVA-Qwen2.5-1.5B-FRFR, heretic_EVA-Qwen2.5-1.5B-v0.0, heretic_Qwen2.5-1.5B-MegaScience, heretic_Qwen2.5-3B, heretic_Qwen2.5-3B-Model-Stock-v2, Heretic-Dolphin3.0-Qwen2.5-1.5B, Heretic-Dolphin3.0-Qwen2.5-3b, Heretic-Qwen2.5-0.5b-RBase, Qwen2.5-1.5B, Qwen2.5-1.5B-Instruct-Heretic, Qwen2.5-14B-Instruct-Heretic, Qwen2.5-32B-Instruct-heretic-GER-GGUF, Qwen2.5-32B-Instruct-heretic-LoRA, Qwen2.5-3B-Instruct-heretic, Qwen2.5-7B-Instruct-1M, Qwen2.5-7B-Instruct-abliterated-v2, Qwen2.5-7B-Instruct-heretic, Qwen2.5-Coder-14B-Instruct, qwen2.5-coder-32b-heretic-swe-sft, Qwen2.5-Coder-32B-Instruct, qwen2.5-coder-32b-instruct-heretic-sft — conversions exist but their publishers did not report KL and refusal numbers, so they are not ranked.