xAI Grok 3.5 beta misses projected launch date

xAI has missed the first projected launch date for the Grok 3.5 model, as it is expected to bring major improvements over Grok 3 and previously released models.

A week ago, Elon Musk, founder of xAI, announced that model 3.5 will be released as early beta to SuperGrok subscribers. The week is about to end, but the company has not shared any details about the new Grok version so far. Such a delay could be related to the model reasoning and preparations for public availability.

Meanwhile, we come to know that 3.5 will focus on improving answers. Large language models (LLMs) are often trained on data available on the internet. The company scraps the online data and trains the model to reply with a plagiarized text. This data in responses can be verified in replies with the source links cited in the conversation.

However, xAI is trying to change its model training with Grok 3.5, while producing more comprehendible replies.

xAI founder Elon Musk revealed Grok 3.5 release date

Musk claimed that Grok is reasoning from first principles and coming with answers that aren’t available on the internet. If this is true, xAI may have a leap forward against its competitors, which still heavily relies on the scrapped training data.

For example, Grok 3.5 could tackle a question like “How does gravity work?”. You can get these answers from the current Grok version, but the new model will help you with the most fundamental concepts of spacetime and mass, constructing an explanation based on general relativity rather than a textbook definition. The same goes for the complicated question such “How a Raptor engine works?”.

The test model is expected to land for SuperGrok subscribers, but we’ll have to wait for more information from the official sources.