pull down to refresh

15 sats \ 1 reply \ @CruncherDefi 26 May \ on: My first game built entirely with a local LLM: 🐍⏪ Rewind Snake AI

How many tokens/second did you achieve on this setup? How does such local model compares to codex/claude?

write
compose
15 sats \ 0 replies \ @rolznz OP 26 May

I get 25TPS. Codex/Claude is significantly better, but I was happily surprised how well this works without relying on one of these centralised companies. Now I’m very much looking forward to the next generation of local models

reply