L3 8B Lunaris v1
Sao10K · 2024-06-26 · 8B parameters
L3-8B-Lunaris-v1 is Sao10K's generalist and role-playing Llama 3 8B model merge, released on Hugging Face under the Llama 3 license. Its published MergeKit recipe uses the TIES merge method to combine Meta-Llama-3-8B-Instruct with four role-playing, storytelling, and general-knowledge-oriented checkpoints, with the goal of balancing creativity and logic.
Benchmark scores
| Benchmark | Score |
|---|---|
| SnakeBench Average Apples | 0.2 |
| SnakeBench Best Apples | 1.0 |
| SnakeBench Rating | 13.8 |
| SnakeBench Total Apples | 5.0 |
| SnakeBench Win Rate | 40.7 |