macOS app configs suggest GPT-5 Auto and Reasoning modes support

OpenAI’s upcoming ChatGPT update, potentially introducing GPT-5, is gathering momentum with new references surfacing across different platforms. The most recent discovery points to a MacOS app configuration file referencing both “GPT-5 Auto” and “GPT-5 Reasoning.” This suggests the possibility of users being able to select between distinct model modes, although the company’s broader strategy has been to eventually simplify model choices for users, possibly phasing out the drop-down selector as model routing matures.

Several credible appearances of GPT-5 have emerged: its mention in a GitHub repository tied to a biological benchmark, a reference in Microsoft Compiler’s Smart Mode, a sighting in an employee’s Cursor screenshot, and now in a MacOS configuration file. These leaks collectively hint at internal tests and restricted rollouts, likely for trusted testers or early partners, rather than full public access. The presence of “Auto” and “Reasoning” modes aligns with past statements from OpenAI leaders, who indicated the next-gen model would function as a router, automatically choosing between reasoning-heavy and general-purpose tasks to cover more user scenarios.

Performance data for suspected GPT-5 variants has shown up on LLM benchmarking arenas, where these models seem to outperform earlier versions in a variety of tasks. For users, this update could mean a shift toward more seamless, context-aware responses, with less need to manually select different model strengths for specific tasks.

The timeline remains unclear, but the pattern of leaks and the volume of speculation—including YouTube stream schedules and connections to the August 5 date - point toward a major release within the next two weeks. It’s also notable that Google’s Tokyo Next event is slated for August 5, which could set up a direct comparison between leading AI labs. If the rollout mirrors past OpenAI launches, initial access could be staggered, focusing on scaling and stability before the selector is removed entirely.

For OpenAI, unifying their model offering would streamline both user experience and backend deployment. This approach fits with the company’s ongoing product consolidation, aiming to make advanced AI tools more broadly accessible while reducing friction for non-technical users. If GPT-5 does launch as described, early adopters—developers, AI researchers, and enterprise teams—are likely to be the first to benefit from these capabilities, with broader consumer access to follow as scaling progresses.