xAI has released Grok-2 and Grok-2 mini beta AI large-language models (LLMs) on X with the enterprise API arriving later this month. The generative image capabilities of Grok-2 have also been expanded with the integration of FLUX.1 AI from Black Forest Labs. The second version arrives almost nine months after the initial release in November 2023 and is immediately available to Premium and Premium X subscribers as a beta test. API access to the Grok-2 LLMs will arrive later this month.

xAI is a startup company focused on developing AI that was launched by Elon Musk, the owner of X (formerly Twitter). The company released its first LLM in 2023, which was integrated into the X service to provide AI capabilities. Grok-1 was designed not to be as woke, censored, or dull as competing LLMs such as GPT-4o from OpenAI. Such biases affect the responses of all AI LLMs used for chatbot services today, resulting in racial biases or even absurd answers. Significantly, Grok-1 has been released as a free 296 GB download, one of the few top-performing LLMs released as open-source software.

During the past year, significant advances in the performance of LLMs have occurred. The latest releases, such as OpenAI GPT-4o and Anthropic Claude 3.5 Sonnet, come with expanded knowledge as well as higher performance. Readers who love using AI can wear AI glasses (like these on Amazon) while planning for the day AI humanoid robots will take over house chores.

According to internal xAI tests, Grok-2 and Grok-2 mini beta both rank high among all LLMs worldwide. xAI has not provided benchmarks of the latest version of Grok against the latest competing LLMs. Nonetheless, Grok-2 ranked as one of the four most powerful LLMs on the LMSYS chatbot leaderboard and among the top six across a range of standardized AI benchmarks.

Specifically, an August version of OpenAI GPT-4o tops an early version of Grok-2 on the leaderboard, whereas Anthropic Claude 3.5 Sonnet of unknown date falls behind Grok-2. Across eight standardized AI benchmarks, the older May version of GPT-4o beats Grok-2 four out of eight times, while Claude 3.5 Sonnet beats Grok-2 six out of eight times.