Elon Musk’s AI, built into X (Twitter). Access to real-time information, a distinctive personality, and deep integration with one of the world’s largest social platforms. History, how to use it, prompts, and technical depth. Official sources only.
Grok is an AI assistant made by xAI, a company founded by Elon Musk. It is built into X (formerly Twitter) — meaning if you have an X account (formerly Twitter), you can access Grok directly within the app. It also has a standalone website at grok.com.
Grok stands out from other AI assistants in two ways: it has access to real-time posts on X, giving it information about events happening right now; and it has a more irreverent, witty personality than most AI tools.
If you already use X (Twitter) regularly and want an AI that knows what is trending, what people are saying about current events, and can engage in more casual conversation — Grok is the natural choice. It lives where you already are.
xAI was founded by Elon Musk in July 2023, months after he left OpenAI’s board and publicly criticised the direction of AI development at other companies. Musk had co-founded OpenAI in 2015 and departed from its board in 2018. His stated reasons for founding xAI were to create an AI focused on “maximum truth-seeking” and to provide an alternative to what he described as overly cautious, politically biased AI systems.
Grok 1.0 launched to premium X subscribers in November 2023. The model was later open-sourced — Grok-1’s weights were released publicly in March 2024, making it one of the largest open-weight models released at the time (314 billion parameters).
Grok-1 was a 314-billion parameter mixture-of-experts (MoE) model. MoE architectures route each input token through a subset of “expert” sub-networks rather than the full parameter set — achieving high effective capacity at lower inference cost than dense models of equivalent size. Grok-1’s weights were released under Apache 2.0 at github.com/xai-org/grok-1.
Grok-2 and Grok-3 architectures have not been fully disclosed. xAI’s Colossus supercomputer — built in Memphis, Tennessee, reportedly with 100,000 NVIDIA H100 GPUs in its first phase — was used to train Grok-3.
xAI (2024). Grok-1 model weights release. github.com/xai-org/grok-1
xAI (2024). “Grok-1 release.” x.ai/blog/grok-os
xAI (2024). “Announcing Grok-2.” x.ai/blog/grok-2
Grok has access to X’s full post stream — the so-called “firehose” — giving it real-time information that no other major AI assistant has. This is a genuine differentiator: while other models have knowledge cutoffs and require web search to access current information, Grok’s X integration is architecturally native.
from openai import OpenAI # xAI uses OpenAI-compatible API
client = OpenAI(
api_key="your-xai-api-key",
base_url="https://api.x.ai/v1"
)
response = client.chat.completions.create(
model="grok-3",
messages=[{"role": "user", "content": "What is trending on X today?"}]
)
print(response.choices[0].message.content)
Full documentation: docs.x.ai