The new artificial intelligence model from xAI did not arrive on time, following a trend.

The list of artificial intelligence models that have missed their launch deadlines continues to expand. Last summer, Elon Musk, founder and CEO of xAI, mentioned that Grok 3, the company’s next major model, would arrive by the end of 2024. Grok, positioned as an alternative to well-known models like OpenAI's GPT-4 and Google's Gemini, has the capability to analyze images and answer questions, in addition to being responsible for various functions on X, Musk's social media platform.

Musk claimed in a July post that "Grok 3, after being trained with 100,000 H100s, should really be something special." In December, he reiterated that this model would represent a significant advancement. However, by January, Grok 3 had still not been launched, and there are no signs that its implementation is near. Codes on the xAI site, detected by informant Tibor Blaho, suggest that an intermediate model named "Grok 2.5" might be released sooner.

This is not the first instance where Musk has set ambitious goals that go unfulfilled. His track record indicates that his announcements about launch timelines are often, at best, unrealistic. In an interview with podcaster Lex Fridman last August, Musk mentioned that Grok 3 would be "hopefully" available in 2024 "if we are lucky." However, the absence of Grok 3 highlights a growing trend in the field of artificial intelligence.

Last year, the startup Anthropic also faced difficulties by failing to launch a successor to its Claude 3 Opus model, later announcing that the Claude 3.5 Opus model would not come out as planned. Reports suggest that while Anthropic completed training on Claude 3.5 Opus last year, they decided it didn't make economic sense to release it. Recently, Google and OpenAI have encountered similar setbacks with their flagship models, which could indicate the limitations of current scaling laws in artificial intelligence.

In the past, it was common to achieve significant performance improvements in models by using large amounts of computational power and broader datasets. However, the advancements of each generation of models have begun to diminish, prompting companies to seek alternative techniques. Musk mentioned in his interview with Fridman the pressure to achieve a state-of-the-art milestone with Grok 3, acknowledging that they might fail to meet that objective.

Furthermore, the delay of Grok 3 could be due to xAI having a much smaller team compared to many of its competitors. Nevertheless, the delay in launch timelines adds to the evidence that traditional training approaches in AI are facing significant challenges.