xAI is currently testing Grok 4.20 in preparation to compete with GPT-5, with a potential launch scheduled for this month to enhance SEO.

Elon Musk-owned xAI is currently testing Grok 4.20, an updated model of Grok 4, which has already shown competitive performance against GPT-5 in various benchmarks, including ARC-AGI 2. While GPT-5 is recognised as one of the leading models for coding and competes directly with Claude Opus 4.1 in coding challenges, Grok has struggled to build comprehensive applications. However, this may soon change with the anticipated release of Grok 4.20. In a recent post on X, Musk hinted at a potential launch for late August.

Additionally, Musk has confirmed that the V7 foundation model has completed its pre-training phase. He stated that the upcoming model would be “natively multimodal,” allowing it to process video and audio streams directly, without the need for conversion. This advancement is expected to enable the model to grasp the nuances of speech that convey mood and emphasis. Although xAI has not officially committed to a release date for Grok 4.20, the likelihood of a significant announcement from the company in the near future appears high. 

Categories: AI Model Development, Competitive Benchmarking, Multimodal Capabilities 

Tags: xAI, Grok 4.20, Model Update, GPT-5, Coding, Multimodal, Video, Audio, Passwords, Data Exfiltration