Perplexity AI Investor Presentation Deck

Perplexity vs rest swyx (in/) @swyx Racing Llamas with GodMode: - @LeptonAl Llama - @perplexity al Llama-13B - @togethercompute Llama-138 - local @TheBlokeAl quantized Llama-13B Findings: Lepton serves its llama model 2x faster than Perplexity, but doesnt specify the size. Given that it answers me with the US Constitution instead of the Declaration, I wonder if it is Llama 7B, in which case we are not comparing apples and oranges. Anyway, have at it! We could be adding Time To First Token and Tokens/s metrics to GodMode soon if the people want it 1:15 E

View entire presentation