According to monitoring by Beating, OpenAI’s unreleased GPT-5.6 model has been successfully accessed by external developers through ChatGPT Pro OAuth authentication in the Codex environment, just three weeks after GPT-5.5’s launch. Probe testing confirmed a context window of 1.5M tokens, a 43% increase from GPT-5.5’s 1.05M token limit. Last week access requests returned “model not supported” errors; this week Pro OAuth credentials granted direct access.
Developer Haider analyzed OpenAI’s release cadence—shrinking from annual to 30-45 day intervals—predicting GPT-5.6 launch in early June. Polymarket currently prices the probability of GPT-5.6 release by June 30 at approximately 85%.
Related News
Sam Altman Announces: OpenAI Codex Grants Companies Two Months of “Free” Use
AI suite supply chain sees two-way attacks: Mistral and fake OpenAI models are both compromised
Meta Tests @meta.ai Assistant on Threads in Five Countries