Xai’s Grok API: new features | Festina Lente - Your leading source of AI news | Turtles AI

As of August, xAI, the AI company founded by Elon Musk, introduced the Grok API, its flagship model. Currently, the API offers the “grok-beta” model, with specific costs for input and output tokens. Despite some initial problems, Grok is integrating advanced features on X and aims to compete with other industry leaders.

Key points:

• xAI launched the Grok API, with only one model available, “grok-beta.”
• API costs are $5 per million input tokens and $15 per million output tokens.
• Grok stands out for its ability to address controversial topics and generate visual content.
• The company is working to improve its functionality and integrate data from other Musk enterprises.

In August, xAI, the AI company founded by Elon Musk, announced the availability of Grok’s API, its generative AI model, accessible through X. Currently, the API features only one model, called “grok-beta,” which costs $5 per million input tokens, equivalent to about 750,000 words, and $15 per million output tokens. The tokens represent segments of data, such as the syllables that make up words. Regarding the nature of the “grok-beta” model, the situation remains unclear; the latest model released is Grok 2, and the documentation also mentions Grok mini, an optimized, lower-cost version. Some users have reported difficulties with payments associated with using the API, highlighting possible technical problems.

The xAI API includes function call functionality, allowing Grok models to be linked to external tools such as databases and search engines, although at present such tools appear not to be fully active. The documentation also suggests the development of models that can process text and images, thus expanding the potential of the API. Musk launched xAI in 2022 and brought the company into the premises previously occupied by OpenAI, where he released Grok for premium X users, who pay a monthly fee of $16.

Grok was described by Musk as having a “rebellious streak,” capable of answering uncomfortable questions that other AI systems tend to ignore. In particular, Grok has no qualms about using strong language, in stark contrast to the more cautious approach of other models. Through an integration with the open source image generator Flux, Grok can also create images on X without restrictions, sparking debates about the need for controls. The model can summarize news and current events, although there is no shortage of errors in its analysis. In the future, Grok may improve X’s search capabilities, as well as post analysis and user biographies.

xAI is trying to compete with giants such as OpenAI and Anthropic in the race for generative AI. In May, the company raised $6 billion in a funding round with the participation of well-known investors. Musk has frequently said that the data collected by X provides xAI with a competitive advantage. Recently, X changed its privacy policies to allow third parties, including xAI, to use post data for model training.

As part of the $6 billion funding, xAI has outlined a plan to train its models using data from Musk’s various associated companies, such as Tesla and SpaceX, suggesting that this could strengthen the technologies in all of these entrepreneurial entities. However, not all Tesla shareholders are in favor of this approach; some have filed lawsuits against Musk, accusing him of diverting resources and talent from Tesla to xAI, perceived as a competing venture.

Musk said xAI is training the next generation of Grok models at a data center in Memphis that has been the subject of controversy over alleged environmental impacts. The company plans to upgrade the data center in the coming year, but that will require approval from the Tennessee Valley Authority. In this complex landscape, Grok continues to evolve in the context of the X ecosystem and AI.

The growing interplay between Musk’s technologies and AI is fueling a broad debate about ethics, innovation, and regulation in the industry.