Particle.news
Download on the App Store
38 ARTICLES
·
last wk.

OpenAI Restores Legacy Models and Unveils Three-Tier Controls for GPT-5

OpenAI aims to balance model availability with scarce compute capacity through restoring GPT-4o for paid subscribers alongside the rollout of Auto, Fast, Thinking modes for GPT-5

Image
Image
Sam Altman, CEO of OpenAI looking pensive off-camera
Image

Overview

  • OpenAI has reinstated GPT-4o and other legacy AI models in the ChatGPT picker for all paid subscribers while keeping GPT-4.5 exclusive to Pro users due to its high GPU cost.
  • GPT-5 now offers selectable Auto, Fast and Thinking modes with Thinking capped at 3,000 messages per week and supporting a 196,000-token context window; Team-plan users face a lower 200-message cap for Thinking.
  • OpenAI outlined a compute prioritization plan that will deliver more usage to current paying ChatGPT users, honor existing API commitments, support roughly 30% new API growth and double its compute fleet over the next five months.
  • Paid subscribers have gained new third-party connectors—including Dropbox, Microsoft Teams, Notion, GitHub, Box, Canva and HubSpot—with some regional restrictions in Europe, Switzerland and the United Kingdom.
  • Sam Altman said OpenAI will update GPT-5’s personality to feel warmer than its current tone and is pursuing more granular per-user customization options following user feedback.