OpenAi has announced the launch of its open source artificial intelligence model (AI) was postponed again. On Saturday, the company’s CEO, Sam Altman, said that due to the additional time required to perform security tests, the release of the AI model was delayed indefinitely. This is the second delay, after the company postponed the statement in June. It is said that the first launch of the São Francisco -based AI since 2019 and the first model of AI model focused on reasoning. In April, the company shared a “open model feedback” form to understand what developers wanted to do with the model.
The release of OpenAi’s open source AI model delayed due to safety tests
In one publish In X (previously known as Twitter), Altman said: “We plan to launch our open weight model next week. We are delaying it; we need time to perform additional safety tests and review high-risk areas. We are still not sure how long it will take us.”
The announcement occurs at a time when OpenAi is also preparing for the launch of the Next generation of its AI FRONTIER model, GPT-5. In recent months, their rivals, such as Google, Anthropic, and Alibaba, have launched the main class AI models that have overcome the latest OpenAi models in several benchmarks.
In addition, these companies (besides Meta and Xai) also made significant contributions to the open community. Alibaba was the latest to enter the trend, because of open origin of the company’s Ernie models.
OpenAi’s planned open source contribution was probably a strategic change to position itself as a leader in this space between companies and consumers (with the proprietary model of GPT-5) and developers (with the open source model). Neither has arrived so far.
According to a techcrunch reportThe IA company plans to launch an AI model that is comparable to its great La -La -LLMS (LLMS) models. However, the company is just planning to release the weights of the model for the open community.
This means that developers and enthusiasts will be able to download the model, adjust their parameters and run it locally. However, they will not be able to learn about their methods and training architecture to replicate the construction of a similar model or build on it.