Did a bit of my own benching on 12 models and 15 questions
Not as extensive as yours ofc and mostly featuring 7B models as the largest size that fits on my 8Gb card is 10.7b
But I think it shows some models I couldn't find on yours, at least at the place in ranking I believe they belong.
qwen's like the most brainwashed model there
Not as extensive as yours ofc and mostly featuring 7B models as the largest size that fits on my 8Gb card is 10.7b
But I think it shows some models I couldn't find on yours, at least at the place in ranking I believe they belong.qwen's like the most brainwashed model there
Check out Liberated Qwen from Ababcus
https://huggingface.co/abacusai
Mistral 0.3 is ruined by the way
doing another type of leaderboard
https://wikifreedia.xyz/based-llm-leaderboard/npub1nlk894teh248w2heuu0x8z6jjg2hyxkwdc8cxgrjtm9lnamlskcsghjm9c
what do you think