Overview
- OpenAI announced on Tuesday that combined active users of Codex and ChatGPT Work have hit 8 million after folding Codex into the ChatGPT desktop app and rolling out GPT‑5.6 Sol.
- GPT‑5.6, launched on July 9, drove a rapid spike in demand that pushed users from about 6 million to 7 million and then to 8 million within days, roughly doubling OpenAI’s prior traffic peak in 48 hours.
- To cope with the load the company reset usage limits, removed a prior five‑hour cap for paid plans, and implemented engineering fixes that include inference optimizations, a reduced context window from 372,000 to 272,000 tokens, and a rollback of experimental high‑reasoning settings.
- About 20% of Codex users are non‑developers using the tools for knowledge work, and firms such as Nvidia and Cisco have already integrated Codex into workflows; flagship Sol access is gated to Plus, Pro, Business and Enterprise plans.
- Rivals reacted quickly — Anthropic extended promotional pricing and raised code‑model limits — and OpenAI warned users to expect possible short‑term capacity hiccups as it adds infrastructure, with broader implications for cloud GPU demand, enterprise compliance needs, and how agentic AI will scale outside developer circles.