Evercore ISI senior managing director Mark Mahaney said on CNBC that OpenAI could debut GPT-6 before the end of 2025. He attributed the timeline to a recent conversation with technology investor Brad Gerstner, who is an OpenAI backer.
The forecast arrives as attention remains fixed on how quickly the company can move from GPT-5 to its next step and whether that advance meaningfully improves reasoning, reliability, and the ability to follow complex instructions.
OpenAI has not provided an official release window, and the company typically keeps road maps close to the vest, so investors should treat the call as informed speculation rather than a formal guide.
The signal matters because model launches have turned into catalysts for the broader artificial intelligence trade. Each major iteration tends to pull spend toward cloud providers, chipmakers, and software platforms that can distribute and monetize new capabilities.
If OpenAI is on track for a late 2025 release, that window roughly aligns with enterprise budgeting cycles for 2026, when many CIOs will be deciding where to place larger bets on AI assistants, copilots, and autonomous agents.
A perceived step change in model quality could also raise the bar for rivals and prompt fresh price competition in inference and API access.
Mahaney’s comment follows months of debate about how much headroom remains in current architectures and whether further gains will depend more on training data, specialized hardware, or new techniques.
The company’s next model will be judged on practical outcomes such as fewer hallucinations, better tool use, and greater consistency across long multi-step tasks.
More Read
Those are the features enterprises care about when evaluating whether to put AI into workflows that touch customer data or revenue. Improvements on those fronts could widen adoption in financial services, health care, and government, where risk controls have slowed pilots.
Apple is expanding its own ambitions, with Apple to launch AI search engine for Siri and Safari in 2026, highlighting how consumer platforms are baking generative features deeply into their products.
Google continues to advance large multimodal models inside Search and Workspace. Meta is seeding assistants across WhatsApp and Instagram.
Each player is trying to capture developer mindshare and distribution, which can shape where inference spending flows long before any specific model wins on benchmarks.
Even if GPT-6 lands in late 2025, integration work, procurement cycles, and risk assessments could push meaningful revenue impact into 2026.
The labor and productivity conversation will run in parallel. The prospect of more capable systems has already intensified discussion around reskilling and job design.
More Read
The debate over AI taking over jobs is likely to sharpen if GPT-6 shows measurable gains in reasoning and autonomy, particularly in office roles that involve research, analysis, and routine documentation.
