This week, several users posted on the X platform that the performance of ChatGPT and OpenAI's programming tool Codex has changed: the responses are stronger and the generated results are more complete, but the waiting time has also increased significantly. Based on these phenomena, outsiders have begun to speculate that OpenAI is conducting small-scale testing of GPT-5.6 within ChatGPT, and some users who chose GPT-5.5 Pro may have been switched to the new model.
Some developers posted comparison results, claiming that the new version performs better in web design, 3D game generation, and code understanding. Others mentioned that, with the same prompts, the time required for the system to generate complete results has increased from about 10 minutes to 20 minutes, 40 minutes, or even more than 1 hour.
Design and 3D tasks are said to have improved.
Developer Anshu Chimala released a comparison video showcasing the landing pages generated under the same prompt, suggesting he may have been using GPT-5.6 Pro ahead of time. Another developer, Dobroslav Radosavljevič, also stated that the model currently running on Codex "feels very different from 5.5."
Conor Dart tested a task that generated a single-word 3D browser game, including physics and camera control. He said the complete generation process took 60 minutes and 15 seconds, significantly slower than the GPT-5.5 Pro's usual response time of about 10 minutes, but the result still surprised him.
Long response time is a common feature
The most consistent signal at present is not a change in the interface, but an increase in processing time. Multiple users have stated that the suspected new model takes a long time to reason on complex tasks, exhibiting the slow response phenomenon commonly seen in earlier high-order models.
However, not all comparisons support the conclusion of a "significant upgrade." An AI benchmarker on the X platform stated that in the same spacecraft modeling task, a version suspected to be GPT-5.6 Pro ran for 87 minutes, while GPT-5.5 Extra High took 34 minutes and 42 seconds. In terms of core geometry, Anthropic's Fable 5 still outperforms both.
OpenAI has not yet confirmed a release date.
Further claims circulating include that the knowledge deadline may be pushed back to December 2025, some test parameters have been increased, and SVG and 3D design capabilities have been enhanced. However, none of this information has been confirmed by OpenAI.
An AI blogger, citing unnamed sources, claims that OpenAI may be secretly distributing the model to select Pro users, with a public release potentially scheduled for June 25th. According to The Information, OpenAI's chief scientist, Jakub Pachocki, told employees that the next-generation model would offer "meaningful improvements" compared to GPT-5.5.
Increased competition drives up market attention
The market's attention to GPT-5.6 is also related to the recent intensification of competition among large-scale models. The report mentions that the Chinese open-source model GLM-5.2 has approached Anthropic's Claude Opus 4.8 and surpassed GPT-5.5 in some engineering benchmark tests.
Meanwhile, Anthropic's Mythos 5 and Fable 5 remain affected by a US export control order issued on June 12 and have not yet resumed operation. This creates a gap in the competition among leading models in the market and makes OpenAI's next-generation models attract more attention.
Additional information:According to The Wall Street Journal, OpenAI is also considering lowering the token pricing for developers and enterprise clients. The report also mentioned that contracts for GPT-5.6 issued on Polymarket between June 22nd and 28th were priced as high as 89% this week.
