This page was generated programmatically; to view the article in its original setting, you can click the link below:
https://www.wired.com/story/openai-o3-reasoning-model-google-gemini/
if you would like to remove this article from our website, please reach out to us
OpenAI has announced an upgraded edition of its most proficient artificial intelligence model ever—one that requires even more time to contemplate questions—just a day subsequent to Google’s introduction of its inaugural model of this kind.
The new model from OpenAI, termed o3, supersedes o1, which was unveiled in September. Similar to o1, the new model dedicates time to mulling over a challenge in order to provide superior responses to inquiries that necessitate a step-by-step logical thought process. (OpenAI decided to bypass the “o2” label, as it is already affiliated with a mobile provider in the UK.)
“We consider this the start of the next era of AI,” stated OpenAI CEO Sam Altman during a livestream on Friday. “You can utilize these models to tackle increasingly sophisticated tasks that demand considerable reasoning.”
According to OpenAI, the o3 model performs significantly better on various metrics than its earlier version, including evaluations that assess advanced coding skills as well as math and science proficiency. It excels threefold over o1 in answering queries presented by ARC-AGI, a standard crafted to evaluate an AI model’s capacity to reason through exceedingly challenging mathematical and logical issues it encounters for the first time.
Google is advancing in a parallel research direction. Noam Shazeer, a researcher at Google, unveiled yesterday in a post on X that the company has developed its own reasoning model, named Gemini 2.0 Flash Thinking. Sundar Pichai, the CEO of Google, referred to it as “our most thoughtful model yet” in his own announcement. The new model from Google achieved a remarkable score on SWE-Bench, a test that assesses a model’s agentic capabilities.
However, OpenAI’s latest o3 model surpasses o1 by 20 percent. “o3 exceeded expectations,” claims Ofir Press, a post-doctoral researcher at Princeton University who contributed to the development of SWE-Bench. “A remarkably surprising improvement; I’m not certain how they accomplished it.”
These two competing models reflect an intensifying rivalry between OpenAI and Google. It is vital for OpenAI to illustrate its capacity for continual progress as it aims to draw in more investment and establish a profitable enterprise. Meanwhile, Google is striving to prove that it continues to lead the AI research sector.
Furthermore, the new models indicate that AI companies are increasingly striving for solutions beyond merely enhancing the size of AI models to extract greater intelligence from them.
This page was generated programmatically; to view the article in its original setting, you can click the link below:
https://www.wired.com/story/openai-o3-reasoning-model-google-gemini/
if you would like to remove this article from our website, please reach out to us
This page has been generated automatically; to view the article in its initial location, you…
This webpage was generated automatically; to view the article in its initial site, you can…
This webpage was generated automatically; to view the article in its original setting, you may…
This page was generated automatically; to view the article in its original form, you can…
This page was generated automatically; to view the article in its initial destination, please follow…
This webpage was generated automatically. To view the article in its initial location, you may…