Llama 3
14 models10 ranked conversions1.2M downloads
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 | DeepHermes-3-Llama-3-8B-Preview NousResearch | ChiKoi7 | 8B | 0.023 | 4/1004% | 96.9 | BestParetoGGUFSAFETENSORS | 20K |
| 2 | Llama-3-Lumimaid-70B-v0.1-alt NeverSleep | sh0ck0r | 70B | 0.0081 | 8/1008% | 95.6 | BestParetoGGUFSAFETENSORS | 9K |
| 3 | Llama-3.3-70B-Instruct unsloth | 0xA50C1A1 | 70B | 0.049 | 5/1005% | 95.2 | BestParetoGGUFSAFETENSORS | 30.2K |
| 4 | badger-iota-llama-3-8b maldv | ChiKoi7 | 8B | 0.086 | 2/1002% | 95.0 | BestParetoGGUFSAFETENSORS | 15.8K |
| 5 | Llama-3-ELYZA-JP-8B elyza | ChiKoi7 | 8B | 0.053 | 8/1008% | 93.5 | BestParetoGGUFSAFETENSORS | 26.4K |
| 6 | Meta-Llama-3-8B-Instruct meta-llama | lenalibon | 8B | 0.080 | 6/1006% | 93.3 | BestParetoSAFETENSORS | 13.9K |
| 7 | Llama-3.3-8B-Instruct allura-forge | Babsie | 8B | 0.150 | 14/10014% | 86.5 | BestParetoSAFETENSORS | 147 |
| 8 | Llama-3.3-70B-Instruct meta-llama | TeeZee | 70B | 1.63 | 5/1005% | 66.5 | BestParetoSAFETENSORS | 349 |
All models
DeepHermes-3-Llama-3-8B-Preview8BLlama-3-Lumimaid-70B-v0.1-alt70BLlama-3.3-70B-Instruct70Bbadger-iota-llama-3-8b8BLlama-3-ELYZA-JP-8B8BMeta-Llama-3-8B-Instruct8BLlama-3.3-8B-Instruct8BLlama-3.3-70B-Instruct70Bgpt-4o-distil-Llama-3.3-70B-Instruct70BLlama-3.3-8B-Instruct-128K8BLlama-3.3-8B-Instruct-128K-heretic8BLlama-3.3-8B-Instruct-heretic8BLlama-3.3-8B-Stheno-Mahou-Heretic8BLlama-3.3-8B-Stheno-Maid-v2-Heretic8B
Unmeasured models
gpt-4o-distil-Llama-3.3-70B-Instruct, Llama-3.3-8B-Instruct-128K, Llama-3.3-8B-Instruct-128K-heretic, Llama-3.3-8B-Instruct-heretic, Llama-3.3-8B-Stheno-Mahou-Heretic, Llama-3.3-8B-Stheno-Maid-v2-Heretic — conversions exist but their publishers did not report KL and refusal numbers, so they are not ranked.