OpenAI Will Release Open-Weight Language Model In The Next Months

JAKARTA OpenAI plans to release the first open-weight language model globally with reasoning capabilities since GPT-2 in the next few months. This was said by OpenAI CEO Sam Altman, on Monday, March 31.

Open-weight language models allow public access to the parameters or weights that have been trained, so developers can analyze and refine models for specific tasks without requiring original training data. This differs from open-source models that provide full access to source codes, training data, and development methodologies.

Altman revealed on platform X that the company will discuss with developers on how to make open-weight models more useful. The first event will be held in San Francisco in the next few weeks, followed by sessions in Europe and Asia-Pacific.

"We still have to make some decisions, so we hold an event for developers to collect input and then try out the initial prototype," said Altman, quoted by VOI from Reuters.

In February, Altman said OpenAI would simplify their AI products while designing a roadmap for the latest models.

Meanwhile, the Microsoft-backed company will have to switch to a for-profit entity before the end of the year to secure a full USD 40 billion funding led by SoftBank Group, according to sources familiar with the matter.

OpenAI previously closed a funding round of $6.6 billion in October last year, and stated that a transition to for-profit companies was needed to acquire sufficient capital to develop the best AI model.