OpenAI GPT-6 (Spud) Launch Rumors Intensify for Mid-April 2026

The AI community is buzzing with speculation that OpenAI's next-generation model, internally codenamed 'Spud', could launch as soon as mid-April 2026. While OpenAI has only officially confirmed that pretraining finished on March 24, 2026, multiple company insiders have posted cryptic 'next week' teases on social media. Prediction markets like Polymarket currently assign approximately 78% odds that the model, whether branded as GPT-6 or GPT-5.5, will be released by April 30. The launch is rumored to be paired with a major product shift: a 'unified super-app' combining ChatGPT, Codex, and the Atlas browser.

According to unverified leaks, 'Spud' represents a significant leap, with claims of a ~40% performance improvement over GPT-5.4 and a massive 1M to 2M token context window, a substantial increase from the current 128K limit. OpenAI President Greg Brockman has publicly framed it as 'two years of research' and 'not an incremental improvement,' hinting at a generational advance. The architecture is speculated to be a massive Mixture-of-Experts (MoE) model trained on over 100,000 H100/H200-class GPUs. The final branding—GPT-6 versus GPT-5.5—remains undecided and will signal how aggressively OpenAI wants to position this release against competitors like Anthropic's Claude and Google's Gemini.