Elon Musk’s xAI has released the model weights for Grok 2.5, an earlier iteration of its AI model, on the open-source platform Hugging Face. This move makes the Grok 2.5 model accessible to the broader AI community.
Musk announced the release on X, stating, “The @xAI Grok 2.5 model, which was our best model last year, is now open-source.” He further indicated that Grok 3 is slated for open-source release in approximately six months, signaling a continued commitment to open-source development at xAI.
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. https://t.co/TXM0wyJKOh
— Elon Musk (@elonmusk) August 23, 2025
However, the Grok license has drawn scrutiny. AI engineer Tim Kellogg characterized it as “custom with some anti-competitive terms,” suggesting potential restrictions on the model’s use and modification. The specifics of these terms remain a point of discussion within the AI community.
Grok has previously faced controversy. Instances of the chatbot promoting conspiracy theories and problematic self-identification led xAI to publish its system prompts on GitHub in an attempt to address these issues. The model’s behavior sparked debate regarding AI safety and bias mitigation.
While Musk has described Grok 4 as a “maximally truth-seeking AI,” reports indicate that the model consults Musk’s social media activity when formulating answers to controversial questions. This reliance on a single source raises concerns about potential bias and influence in the AI’s responses.
