Elon Musk's artificial intelligence company xAI is preparing to release the latest version of its Grok generative AI platform, promising substantial improvements in its performance.
The Grok-1.5 model will come with upgraded reasoning capabilities and a context length of 128,000 tokens, and will be accessed by early testers and existing Grok users, the San Francisco-based company said on Friday.
Tokens in AI are the basic unit to calculate the length of text and can include punctuation marks and spaces, and vary from one language to another. The bigger the token count, the more comprehensive the generated results are.
“This allows Grok to have an increased memory capacity of up to 16 times the previous context length, enabling it to utilise information from substantially longer documents,” xAI said.
Benchmarks presented by xAI claim that Grok-1.5's performance is now nearly at par with generative AI's leading large language models, including Google's Gemini, Anthropic's Claude and industry leader OpenAI's GPT.
Mr Musk said Grok-1.5 will be available on X, the platform formerly known as Twitter that he bought in 2022, next week.
He also hinted that Grok's subsequent version, Grok 2, is “in training now” and “should exceed current AI on all metrics”, without elaborating.
Grok is being positioned by the company as a “conversational AI for understanding the universe” – aligned with Mr Musk's mission statement when he announced in July last year that xAI was meant “to understand reality” and “the true nature of the universe”.
While all generative AI chatbots rely on LLMs, the underlying algorithm that uses deep learning and analyses significant amounts of data to generate content, Grok has one distinct edge, according to Mr Musk – its real-time access to information via the X platform.
The previous version, Grok-1, was released two weeks ago. It was trained on 33 billion parameters, had gone through several iterations over the past four months and is “significantly more powerful”, xAI had said at the time.
Grok-1.5's figures appear to be impressive on paper: it was able to double its score in the maths benchmark, besting Claude 3 Sonnet, according to data presented by xAI.
In the HumanEval benchmark, which measures an LLM's capability to generate code, Grok-1.5 beats out Gemini Pro 1.5 and GPT-4, it added.
“By releasing the model weights and network architecture of Grok-1 two weeks ago, we presented a glimpse into the progress xAI had made up until last November. Since then, we have improved reasoning and problem-solving capabilities in our latest model,” xAI said.
Grok is only available to subscribers of X's Premium+ tier, which starts at $16 per month, or $168 a year. It comes with revenue sharing, gives users a large reply boost, grants access to other creator tools and removes advertising from feeds.
X has not revealed its Premium+ subscriber numbers. However, the latest data from Statista shows that there were about 640,000 subscribers as of April last year. A more recent study from data analyst Travis Brown said there were about 892,000 as of September.
At the time of Grok-1's launch, Mr Musk also said that xAI will make Grok open source, doubling down on his offensive against ChatGPT maker OpenAI.
Mr Musk sued OpenAI and its chief executive Sam Altman on February 29, alleging it abandoned its initial non-profit mission and called the company “a lie”. OpenAI soon fired back, saying that Mr Musk was just regretful he wasn't part of the company's current success.
