OpenAI has officially unveiled GPT-5, the latest iteration of its flagship large language model, marking what the company describes as a "generational leap" in artificial intelligence reasoning capabilities.

The announcement, made at a packed event in San Francisco, showcased the model's ability to tackle complex, multi-step problems that had previously stumped AI systems. In live demonstrations, GPT-5 solved graduate-level mathematics problems, debugged intricate codebases, and conducted nuanced scientific analysis.

"We're seeing emergent capabilities that we didn't explicitly train for," said OpenAI's chief scientist during the presentation. "The model is developing what we can only describe as genuine problem-solving intuition."

Independent benchmarks confirm the improvements. GPT-5 scores 92% on the notoriously difficult ARC-AGI benchmark, up from GPT-4's 68%. On coding tasks, it achieves a 94% pass rate on HumanEval, and demonstrates near-human performance on complex reasoning tasks.

The model also introduces a new "chain of thought" architecture that makes its reasoning process more transparent. Users can now observe the model's step-by-step logic, making it easier to identify errors and understand how conclusions are reached.

Industry analysts predict the release will accelerate AI adoption across sectors. "This isn't just an incremental improvement," noted one technology analyst. "GPT-5 represents a qualitative shift in what AI systems can reliably accomplish."

The model will be available through OpenAI's API starting next month, with consumer access following shortly after. Pricing details have not yet been disclosed, though the company indicated it would maintain competitive rates.