Openai's astra model demonstrates advanced capabilities in game development and code writing

On August 29, testingcatalog reported that OpenAI is accelerating the release of its Astra model, which has entered a new testing phase called mozaik-alpha-fdm, expanding internal testing. Astra, part of OpenAI's next-generation AI series following Sol, Terra, Luna, and named after the Latin word for 'stars,' has shown significant breakthroughs in mathematics, physics, biology, medicine, chemistry, and cybersecurity, with claims of solving long-standing mathematical problems. As of August 2026, Astra has not been officially confirmed as GPT-6, but media and industry sources widely speculate it could be the GPT-6 or a candidate name, potentially with a parameter scale of 10 trillion. Testers reported that Astra can create a game similar to GTA 2 in one interaction, along with detailed websites, 3D objects, and voxel environments. However, testers noted that Astra currently operates in zero-sample testing with Max inference strength, resulting in longer inference times compared to GPT-5.6 Sol. The media suggests that if Astra maintains this efficiency in its final release, it could benefit developers, designers, and entrepreneurs. However, the model's current performance highlights the challenges in achieving such high efficiency in practical applications.