Key facts
- Users are reporting that ChatGPT and Codex outputs have become noticeably sharper and more capable.
- Some users are experiencing significantly longer processing times, suggesting more complex computations.
- Screenshots and videos shared online compare outputs from a suspected GPT-5.6 model against GPT-5.5 Pro.
- Improvements have been noted in areas like web design, 3D game development, and robotic simulations.
- Leaked details suggest a knowledge cutoff extended to December 2025 and enhanced reasoning capabilities.
Users across social media platform X have reported a noticeable increase in the intelligence and capability of ChatGPT and Codex over the past week, sparking rumors of a stealth test for OpenAI's next-generation model, GPT-5.6.
Several developers and AI enthusiasts have shared side-by-side comparisons, showcasing improved outputs in areas such as web design, 3D video game generation, and robotic simulations. These enhanced results are often accompanied by significantly longer processing times, with some tasks taking over an hour compared to the usual 10-minute mark for GPT-5.5 Pro.
Leaked details, attributed to sources like leaker Pankaj Kumar, suggest GPT-5.6 could feature a knowledge cutoff extended to December 2025, an increased reasoning-effort setting, and improved performance in design tasks, potentially surpassing competitors like Anthropic's Fable 5 in certain benchmarks. AI influencer Leo cited unnamed sources indicating that GPT-5.6 is being stealthily tested on some ChatGPT Pro accounts, with a potential public launch rumored for June 25.
While OpenAI has not officially confirmed these developments, Chief Scientist Jakub Pachocki reportedly informed staff that the upcoming model represents a meaningful improvement over GPT-5.5. The timing of these rumors coincides with increased competition in the AI space, including China's open-source GLM-5.2 model and Anthropic's recent models being subject to a U.S. export control directive. OpenAI is also reportedly considering price cuts for its services in anticipation of potential IPOs from both itself and Anthropic.