Is ChatGPT 4o mini enough as LLM or whether full 4o is required?

[alexandre-emmanuel]

ChatGPT 4o mini API is way more cheaper than ChatGPT 4o and I wonder - whether mini i enough as LLM for OpenDevin? What is your experience?

I queried both models for 'Pls, write simple Form component with React, RTK/Redux store.' and they both gave comprehensive, complete, multifile solution.

Out of curiosity I asked also 'Pls, write formalization of inf-category in Coq?' and even the mini model was not afraid of task and gave quite decent and quite full formalization, both model failed to tackle the necessary coherence requirements, but that is fine.

[mamoodi]

I think it might depend on the task at hand? I haven't played with 4o-mini yet but I would assume it would be capable of a lot of tasks/prompts that normally get thrown at it.