摘要：TMTPOST -- OpenAI’s next-generation artificial intelligence model, GPT-5, is nearing release, intensifying anticipation across the

Credit: CFP

TMTPOST -- OpenAI’s next-generation artificial intelligence model, GPT-5, is nearing release, intensifying anticipation across the tech industry over whether the upgrade can match the leap seen in previous iterations of the company’s breakthrough large language models.

The forthcoming model, a successor to GPT-4, is expected to launch in the coming days, according to people familiar with the matter. Early testers told Reuters they were impressed by GPT-5’s ability to code and solve complex math and science problems, though several noted the improvement over GPT-4 may be less dramatic than the jump from GPT-3 to GPT-4.

The feedback underscores growing scrutiny of whether OpenAI — backed by Microsoft Corp. and last valued at around $300 billion — can maintain the pace of progress that helped turn ChatGPT into one of the fastest-growing consumer applications in history.

“OpenAI made such a great leap from GPT-3 to GPT-4, that ever since then, there has been an enormous amount of anticipation over GPT-5,” said Navin Chaddha, managing partner at AI-focused venture capital firm Mayfield. “The hope is that GPT-5 will unlock AI applications that move beyond chat into fully autonomous task execution."

OpenAI has not provided an official release date, but Boris Power, the company’s head of applied research, signaled the launch was imminent in a post on X earlier this week, writing: “Excited to see how the public receives GPT-5.”

GPT-4, released in March 2023, was a milestone in AI capability, surpassing GPT-3.5 by scoring in the top 10% on a simulated bar exam, compared with the bottom 10% for its predecessor. It fueled a wave of adoption and competitive responses from rivals including Alphabet Inc.’s Google, Amazon-backed Anthropic, and Meta Platforms Inc., which released its Llama 3 models within a year.

But as the industry races ahead, OpenAI is facing the limits of its original scaling strategy. GPT-4’s gains were driven by increases in compute power and training data — but both are becoming harder to scale. Former Chief Scientist Ilya Sutskever warned last year that while compute continues to grow, the volume of high-quality human-generated text data is reaching a ceiling.

In addition to data constraints, OpenAI has wrestled with growing complexity in training “frontier” models. Full training runs can take months and are increasingly prone to hardware failures. The final performance of these models often remains uncertain until late in the development process.

To address those challenges, OpenAI is now pursuing a complementary strategy called “test-time compute,” which allocates more processing power to specific tasks — such as math or reasoning — after a model is deployed. CEO Sam Altman has said GPT-5 will combine both test-time compute and large-scale model architecture to push capabilities further.

At the same time, Altman has acknowledged that the company’s portfolio of models and products has grown increasingly complex — a comment widely interpreted as a nod to the fragmented nature of OpenAI’s offerings across ChatGPT, APIs, developer tools, and enterprise integrations.

As OpenAI prepares to release GPT-5, investors, developers, and enterprise clients will be watching closely to see whether the new model can sustain the company’s momentum — and deliver the kind of breakthrough that once again redefines the frontier of artificial intelligence.

来源：钛媒体APP一点号