Summary

  • OpenAI is responding to customer feedback and making GPT-4 the default again for its paying users, alongside the new ‘Show additional models’ setting which offers restores access to other popular LLMs.
  • This reverses a decision made last week when GPT-5 was launched to replace GPT-4 as the default model, despite the fact that many users preferred GPT-4’s outputs.
  • Customers will now be able to choose between the ‘Auto’, ‘Fast’, and ‘Thinking’ modes for GPT-5, with the ‘Thinking’ mode carrying a 3,000 messages-per-week cap after which subscribers can use the lighter ‘GPT-5 Thinking mini’ mode.
  • OpenAI’s CEO Sam Altman also hinted at the possibility of a personality tweak for GPT-5 to make it feel warmer, but less polarising than GPT-4.
  • The changes are an attempt to placate users frustrated by the sudden launch of GPT-5 and the subsequent deprecation of the older models.

By Carl Franzen

Original Article