OpenAI to launch its first ‘open’ language mannequin since GPT-2 in 2019

OpenAI is about to launch an “open” model of its language mannequin this yr, permitting builders to run the mannequin on their very own {hardware}.

In an replace posted to X on March 31, OpenAI CEO Sam Altman said the factitious intelligence agency would launch the highly effective “new open-weight language mannequin with the reasoning” within the coming months however first needed to assemble suggestions about “methods to make it maximally helpful.”

“We’ve been enthusiastic about this for a very long time, however different priorities have taken priority. Now it feels essential to do,” he mentioned, including it was the primary “open-weight” mannequin since GPT-2 in 2019.

Including that: “We nonetheless have some choices to make, so we're internet hosting developer occasions to assemble suggestions and later play with early prototypes.”

An open-weight language mannequin is publicly accessible for anybody to make use of, obtain, modify or deploy for their very own functions. It’s not as open as an open-source mannequin, however it will be a change from GPT-3 and GPT-4, which have been absolutely closed.

Altman mentioned developer occasions for GPT-2 will begin in San Francisco, California — the place OpenAI is headquartered — within the subsequent few weeks, adopted by periods in Europe and the Asia–Pacific area, in accordance with Altman.

“We’re excited to see what builders construct and the way massive corporations and governments use it the place they like to run a mannequin themselves,” Altman mentioned.

“We'll do further work provided that we all know this mannequin can be modified post-release.”

OpenAI’s first “open” mannequin” since GPT-2 in 2019

OpenAI’s open-weight language mannequin GPT-2 was partially launched in February 2019, adopted by a full launch in November of the identical yr.

Associated: ‘Our GPUs are melting’ — OpenAI puts limiter in after Ghibli-tsunami

Altman additionally mentioned on Feb. 12 his firm wants to ship GPT-4.5 and GPT-5 within the coming weeks or months.

The AI arms race has been heating up with the launch of rival DeepSeek, which capabilities equally to ChatGPT however was reportedly created at a fraction of the price and time.

Alibaba Group launched its new open-source AI mannequin for cost-effective AI brokers on March 26, whereas Google introduced Gemini 2.5, its newest experimental synthetic intelligence mannequin, on March 25.

In the meantime, Meta CEO Mark Zuckerberg said in a March 19 Threads submit that the agency’s AI mannequin household, Llama, launched in February 2023, had hit 1 billion downloads.

Journal: ‘Chernobyl’ needed to wake people to AI risks, Studio Ghibli memes: AI Eye