In early 2025, Elon Musk’s Grok AI became one of the most discussed on the market. The AI chatbot, integrated into X (formerly Twitter), has seen exponential user growth in a short period.
Today, Grok refers to a series of AI models developed by xAI, including Grok-1, Grok-1.5, Grok-2, and the latest Grok-3 AI.
In this article, you’ll learn the latest Grok statistics about user growth, visit trends, benchmarks, and other key insights into its expansion in the AI market.
Let’s first explore Grok statistics on the number of visits, user demographic & geography and compare it with other generative AI models.
Grok AI’s website visits surged from 1.2 million users in September 2024 to 25.82 million in February 2025. That’s an increase of over 24.6 million visits in just five months.
Let’s look at the following Grok statistics that show website visits over the last months.
September 2024
1.2M
-
-
October 2024
1.5M
+0.3M
↑25%
November 2024
3.1M
+1.6M
↑106.67%
December 2024
8.05M
+4.9M
↑158.06%
January 2025
4.81M
-3.2M
↓40%
February 2025
(Grok-3 release)
25.82M
+21M
↑436.34%
According to the latest available data, Grok’s popularity saw its most significant spike in February 2025, coinciding with the release of its 3rd version. Grok user growth jumped by 436.34% in a single month. Compared to January, the platform gained 21 million more visits.
Grok’s growth trend follows a steady rise in late 2024. The traffic more than doubled in November (106.67%) and December (158.06%). However, January 2025 saw a temporary decline of 40%.
Source: Semrush
Grok’s user base is predominantly male. Men make up 66.91% of users, while women account for 33.09%.
Grok is most popular among young to middle-aged adults. The largest age group is 25-34 (33.39% of all). Younger users aged 18-24 follow, making up 21.46% of the audience. The 35-44 group accounts for 19.10%, while 12.96% fall within the 45-54 range.
Older demographics are less prevalent: 8.21% are between 55 and 64, and just 4.88% are 65 or older.
Source: Similarweb
Grok’s user base is quite diverse. The United States accounts for 14.61% of all users. India follows with 11.47%, while China holds 9.13%. Vietnam makes up 3.84%. In fact, 60.95% of users come from other countries.
United States
14.61%
India
11.47%
China
9.13%
Vietnam
3.84%
Others (combined)
60.95%
Now, let’s compare Grok stats with other artificial intelligence models.
Despite Grok-3’s launch driving a traffic boost, total visits and session duration still lag behind competitors. Grok’s 25.82 million visits are far fewer than ChatGPT’s 5.19 billion or even Perplexity AI’s 165.92 million.
ChatGPT
5.19B
08:13
Perplexity AI
165.92M
08:46
Gemini
139.37M
06:27
Claude
111.85M
07:46
Grok
25.82M
04:35
Mistral AI
20.11M
08:20
Grok AI’s average visit duration was 4 minutes and 35 seconds, the shortest among competitors. In contrast, Perplexity AI users stayed the longest at 8 minutes and 46 seconds. Even Mistral AI, with fewer visits (20.11 million), had nearly double the session time at 8 minutes and 20 seconds.
Grok-1
Nov 3, 2023
The first version of Grok, developed by xAI in Python and Rust, designed for basic text understanding and response generation. Built on a neural network architecture with 314 billion parameters. Released under the Apache-2.0 license.
Grok-1.5
May 15, 2024
An improvement over Grok-1 with enhanced reasoning capabilities and a context length of 128,000 tokens. Released under a proprietary license.
Grok-2
Aug 14, 2024
Upgraded performance and reasoning over Grok-1.5, with added image & PDF understanding capabilities and web search. Released under a proprietary license.
Grok-3
Feb 17, 2025
The latest version, trained with 10 times more computing power than Grok-2. Features advanced reasoning capabilities similar to OpenAI's o3. Released under a proprietary license.
Since its release, the new Elon Musk chatbot has made significant progress in AI intelligence. As of March 2025, Grok-3 Reasoning outperforms many competitors in scientific reasoning and mathematical problem-solving. However, the standard Grok 3 still lags behind top-tier models in some areas.
Below is a breakdown of Grok performance in key AI benchmarks.
The Artificial Analysis Intelligence Index (AAII) ranks AI models based on reasoning, knowledge, coding, and mathematical abilities. It uses tests like MMLU-Pro, GPQA Diamond, Humanity’s Last Exam, LiveCodeBench, SciCode, AIME, and MATH-500.
The latest version, released in February 2025, shows that Grok-3 Reasoning Beta leads among AI models with 66 points. It ranks higher than ChatGPT o1 (62) and DeepSeek R1 (60).
Standard Grok 3, in turn, scores 51 points – below top-tier competitors but above GPT-4.5 preview (51) and Gemini 2.0 Pro Experimental (49).
According to the Grok statistics above, Elon Musk’s chatbot performs well in key intelligence tests. While Grok-3 Reasoning Beta leads in overall AI rankings, its most competitive results appear in two key evaluations:
Below is a closer look at how Grok performed in these areas compared to other AI models.
GPQA Diamond is a benchmark designed to test an AI model’s ability to analyze complex scientific concepts. It evaluates how well a model understands advanced topics in physics, chemistry, biology, and other scientific fields. Unlike simple multiple-choice exams, GPQA Diamond challenges AI to solve multi-step problems and provide well-reasoned explanations.
Grok-3 Reasoning Beta scored 80% on the GPQA Diamond test, ahead of Claude 3.7 Sonnet Thinking (77%).
Standard Grok 3 follows closely at 75%. It outperforms OpenAI’s GPT-4.5 preview (71%), Gemini 2.0 Pro (62%), and GPT-4o (54%).
Next, the AIME (American Invitational Mathematics Examination) 2024 is a benchmark that assesses an AI model’s ability to solve high-level mathematical problems. It focuses on complex, multi-step reasoning in number theory, geometry, and combinatorics.
Grok-3 Reasoning Beta leads in AIME 2024 again with 84%. OpenAI o1 follows at 72%, with DeepSeek R1 at 68%. Standard Grok-3 falls behind at 52% but still outperforms Claude 3.7 Sonnet Thinking (49%), GPT-4.5 Preview (37%), and Gemini 2.0 Pro (36%).
The context window defines how much information an AI model can process in one interaction. Grok’s usage limit is 1 million tokens, nearly 8x higher than most competitors.
GPT-4o, Llama 3.1+, and DeepSeek each offer only 128K tokens. Claude allows 200K tokens, which is still far below Grok-3’s limit.
Only Gemini (2.0 Pro Experimental and 1.5 Pro) surpasses Grok-3 with a 2 million token limit. But, still, Grok 3 can retain more context and handle longer conversations than the average model.
Gemini 2.0 Pro Experimental
2m
Gemini 1.5 Pro
2m
Grok 3
1m
Nova Pro
300k
Command A
256k
Claude 3.5 Haiku
200k
o1
200k
GPT-4o
128k
GPT-4.5 (Preview)
128k
Llama 3.3
128k
Mistral Large 2
128k
DeepSeek R1
128k
Source: Artificial Analysis
All in all, Grok AI experienced rapid growth in 2025, with traffic surging by 436% following the release of Grok-3. Benchmark tests already highlight its leading positions in scientific reasoning and mathematical problem-solving. And based on current Grok statistics, Elon Musk’s chatbot’s expansion shows no signs of slowing down.
Get a consultation and start building your dream team ASAP.
Start hiringGrok is an AI chatbot developed by xAI, a company founded by Elon Musk. It provides conversational AI capabilities and is integrated into X (formerly Twitter). Grok 3 also has reasoning features, DeepSearch, an Aurora text-to-image generator, web search capabilities, and more.
As of March 2025, Grok-3 is free for all X users. Initially, access was limited to X Premium+ subscribers, but now every user can log into their X account and select Grok from the menu. xAI indicates that this free access is available “until our servers melt,” so it may be temporary.
For comparison, Grok-1 is a 314-billion parameter Mixture-of-Experts model.
Grok accesses and analyzes live posts and news updates from X in real time. Thus, it doesn’t have a knowledge cutoff date and can provide up-to-date responses.
