The SpaceX founder previously said he plans to use Grok 4 to identify, correct, and improve online knowledge.
Highlights
The xAI team, led by Elon Musk, will release the Grok 4 model with improved coding capabilities soon after July 4.
The main goal of Grok 4 is to improve AI model training by properly reviewing human knowledge-based data and increasing inference accuracy.
After a long development phase with his AI team, Elon Musk announced today that xAI’s “Grok 4” model will be released after July 4. The model will have a special coding component that will require “another big run” before release, he said.
The announcement follows the release of Grok 3 in February. xAI touted Grok 3 as having 10 times the computing power of its predecessor.
The current version of Grok is available to X Premium Plus subscribers and on xAI’s dedicated platforms.
Last week, Musk revealed his plans to use the new model to modify existing AI training data.
Grok 4, which has advanced reasoning — will be used to reconstruct all of the human knowledge available online, correcting errors, adding missing information, and removing what Musk calls “garbage data.”
The goal is to retrain the model on this refined dataset. This will allow it to avoid the biased and flawed sources that large language models typically rely on.
xAI, founded by Musk in 2023, developed Grok 3 using its powerful “Colossus” supercomputer. It was built using more than 100,000 hours of Nvidia GPU processing over a nine-month period.
Earlier this month, X announced that it had teamed up with Polymarket as its official prediction partner. By working with Polymarket, a decentralized prediction platform, the goal is to bring market-based predictions into the mainstream, as well as improve the accuracy and transparency of predictive analytics.
For more information: Click Now