Elon Musk has said that his AI startup xAI has open-sourced its Grok 2.5 language model and plans to do the same for its Grok 3 model in the next 6 months. Notably, Grok 3 was launched by Musk and his team in a live-stream in February, and the billionaire has previously stated that he plans to launch the Grok 5 model by the end of this year.
‘The @xAI Grok 2.5 model, which was our best model last year, is now open source. Grok 3 will be made open source in about 6 months.’ Musk wrote in a post on X (formerly Twitter).
Prior to this, xAI had also open-sourced its Grok 1 model last year.
Notably, Musk’s chief rival OpenAI had announced the launch of two new open-source models earlier in the year, marking the first time that the ChatGPT maker released an open-source model since GPT-2.
Among Western AI companies, Meta’s Llama has been one of the pioneers of open-source models; the company has launched all four of its Llama models under open licenses so far.
Meanwhile, Chinese startups like Qwen and DeepSeek had gained prominence earlier in the year owing to their open-source nature and put pressure on their Western counterparts to release models under similar licenses.
Grok AI at the centre of controversy:
Notably, Musk’s Grok AI has been at the centre of controversy recently as it was revealed that the chatbot leaked hundreds of thousands of private conversations of its users online. When users clicked on the ‘share’ button on the chatbot, it led to the creation of a unique URL which also made the conversations available to search engines like Google, Microsoft’s Bing, and DuckDuckGo, putting the privacy of thousands of users at risk.
Grok had also been involved in another controversy this year as it went on to worship Adolf Hitler on X and even claimed the name ‘Mecha Hitler’. xAI later attributed the issue to a deprecated code, but the chatbot continued to embarrass the company as it later went on to answer its surname as Hitler in response to questions to the latest Grok 4 Heavy model.
