GPT-5, the next evolution in AI from OpenAI, is set to achieve Ph.D.-level intelligence in specific tasks by late 2025 or early 2026, according to CTO Mira Murati.
Internally codenamed “Gobi” and “Arrakis,” GPT-5 will be a multimodal model with an expected 52 trillion parameters. While rumors initially suggested a 2023 release, CEO Sam Altman recently emphasized there is no fixed timeline, and the model’s name may change.
The next major model of ChatGPT has been long anticipated, and GPT-5 is set to bring a substantial leap in AI capabilities. OpenAI Chief Technology Officer Mira Murati provided fresh insights into its potential during an interview with Dartmouth Engineering, shared on X.
Murati compared the progression from GPT-4 to GPT-5 to the growth from a high school level to a university level of understanding, indicating a significant enhancement in the model’s sophistication and capabilities.
GPT-5, which has been internally codenamed “Gobi” and “Arrakis” during its development phases, is being closely monitored by industry analysts.
At the 2024 Berggruen Salon in LA, Microsoft CTO Kevin Scott highlighted GPT-5’s potential to pass complex exams, reflecting significant improvements in reasoning and problem-solving. He also noted the decreasing barrier to entry in AI, making powerful tools more accessible to a wider audience. Training for GPT-5 began in December 2023, with a public release anticipated post-2024 US elections.
Alan D. Thompson, AI expert and advisor of Fortune 500 companies, major governments, and UN member states, has created a comprehensive database of large language models. Before recent timeline comments, Thompson predicted a public release of GPT-5 in December 2024, post-US elections.
When asked about the release timeline, Murati indicated that GPT-5 could be expected in late 2025 or early 2026. This is a shift from earlier rumors that suggested a launch as early as late 2023, which were later adjusted to summer 2024.
The delayed timeline might disappoint those eager for the next big advancement. However, Murati’s clarification suggests that the additional time will allow OpenAI to develop a more robust and intelligent system than previously anticipated.
Adding to the discussion, OpenAI CEO Sam Altman recently appeared on the All-In podcast, where he emphasized that there is no fixed timeline for GPT-5. Altman mentioned that the next iteration might not even be called GPT-5, underscoring the fluid nature of the project and the emphasis on quality over adhering to a strict schedule.
The claim by Murati that GPT-5 will achieve “Ph.D.-level” intelligence in certain tasks has been backed up by previous statements from industry leaders.
Microsoft CTO Kevin Scott has suggested that next-gen AI systems will be capable of passing Ph.D. exams, thanks to improvements in memory and reasoning operations. Murati echoes this sentiment, though she emphasizes that such high-level performance will be task-specific.
Current AI systems already exhibit human-level abilities in some areas, and GPT-5 aims to expand these capabilities further, demonstrating advanced reasoning and knowledge application in complex scenarios.