Openai to release the first “open” linguistic model since GPT-2 in 2019

Openai has been assigned to launch a “open” language version this year, allowing developers to run the model on their own devices.

In an update posted on X on March 31, CEO of Openai Sam Altman He said The AI ​​will issue a strong model for an “open -open language model” with thinking “in the coming months, but it first wanted to collect notes on” how to make it useful to the maximum. “

“We have thought about this for a long time, but other priorities have precedence. Now it is important to do it,” he said, adding that it was the first “open” model since GPT-2 in 2019.

He added: “We still have some decisions to be taken, so we host the events of the developers to collect comments and play later with early initial models.”

An open -weight language model is generally available for anyone to use, download, modify, or spread for its own purposes. It is not open like an open source model, but it will be a change from the GPT-3 and GPT-4, which has been completely closed.

Altman said that the events of the developers of GPT-2 will start in San Francisco, California-where Openai is headquarters-in the next few weeks, followed by sessions in Europe, Asia and the Pacific, according to Altman.

“We are excited to find out what the developers are based on, how many companies and major governments are using as they prefer to run a model themselves,” Altman said.

“We will do additional work, given that we know that this model will be modified after the release.”

The first “Open” model for Openai since GPT-2 in 2019

The Openai GPT-2 language model was partially released in February 2019, followed by a full version in November of the same year.

Related to: “Our GPUS”-Openai puts the specified in the Ghibli-Tsunami dimension

Altman also said on February 12, his company wants to ship GPT-4.5 and GPT-5 in the coming weeks or months.

The AI ​​Arms Arm race was heated with the release of the Deepseek competitor, which works similar to ChatgPT but was created in a small part of the cost and time.

Alibaba Group has launched the new open -source AI model for effective technician artificial intelligence agents on March 26, while Google Gemini 2.5 presented the latest model of experimental artificial intelligence, on March 25.

At the same time, Meta CEO Mark Zuckerberg He said In the interconnected publication publication on March 19 that the company’s AI Model family, Llama, which was released in February 2023, recorded a billion download.

magazine: “Chernobyl” needs to awaken people to the dangers