Gemini is the latest iteration of Google’s AI technology. This solution has become a valuable tool for tackling modern business challenges. Use cases include marketing, data analysis, idea generation, report generation, etc.
As of March 2025, Gemini includes Gemini 1.5 (Pro & Frash), Gemini 2.0 (Flash & Flash-Lite), and the latest Gemini 2.5 Pro.
In this article, you’ll find out the detailed Google Gemini statistics. We’ll highlight its strengths and compare its features to those of other leading language models. Let’s explore the numbers to understand how this solution stands out in the AI competition.
Below are the most notable Google Gemini stats:
Let’s go over the key details of this language model launched by Google:
Google Gemini AI release date
Google Gemini: December 6, 2023
Bard (former version): March 21, 2023
Parent company
Google, under Alphabet Inc.
Available regions
Over 239 countries and territories
Language support
Over 46 languages, including Chinese, Korean, Arabic, Hindi, and Spanish
Model variants
Gemini 2.5 Pro Experimental: For enhanced thinking and reasoning, multimodal understanding, advanced coding
Gemini 2.0 Flash: For next-generation features, speed, thinking, real-time streaming, and multimodal generation
Gemini 2.0 Flash-Lite: Cost-efficient and low-latency alternative to Gemini 2.0 Flash
Gemini 1.5 Pro: For complex reasoning tasks, e.g., code and text generation, text editing, problem-solving, etc.
Gemini 1.5 Flash: For fast and versatile performance across a diverse variety of tasks
Supported data types for input
Text, Image, Video, Audio
Knowledge cutoff
January 2025 (Gemini 2.5 Pro)
Supported # tokens for input
1M (Gemini 2.5 Pro)
Websites Like Gemini
ChatGPT, Claude, Grok, DeepSeek
Source: Google AI for Developers, Google DeepMind
Now that we got to know what Google Gemini is, let’s move into the usage stats.
Google Gemini attracts millions of users globally. Let’s review its performance in several key usage trends.
There were 284.1 million monthly visits in February 2025
In February 2025, Google Gemini statistics showed 193.3 million visits on desktop and 90.83 million visits on mobile. Visits peaked this month with 284.1 million total (a 6.14% increase from January 2025).
However, mobile traffic has been declining in recent months. Specifically, there were only 81.72 million mobile visits in January 2025, a 9.88% decrease from 90.69 million mobile visits in December 2024.
In the Claude vs Google Gemini vs ChatGPT comparison, the latter dominates in total visits. In February 2025, ChatGPT had an impressive 3.9 billion total visits. It significantly outpaced Google Gemini’s 284.1 million. Meanwhile, Claude recorded 73 million.
Google Gemini attracted 67.29 million unique visitors in February 2025
In January 2025, the platform had 62.91 million unique visitors overall. There was a slight increase, with unique visitors raising to 67.29 million by April.
Gemini number of users on desktop started at 36.33 million in December 2024 and increased slightly to 40.35 million in February 2025. Meanwhile, mobile unique visitors dropped from 27.33 million to 26.94 million over the same period.
The bounce rate on mobile increased to 44.78% in Fabruary 2025
Google Gemini statistics by month show a relatively stable bounce rate with slight fluctuations. In January 2025, the desktop bounce rate was 30.63%. The value decreased slightly to 27.1% by February 2025. The mobile indicator saw a minor decrease from 45.05% in January to 44.78% in February.
Dec 24
35.34%
–
44.04%
–
Jan 25
30.63%
-13.33%
45.05%
+2.29%
Feb 25
27.1%
-11.52%
44.78%
-0.6%
Gemini users viewed an average of 3.2 pages per visit in February 2025
Engagement on Google Gemini is evident through the pages per visit metric. Desktop users visited an average of 2.92 pages per session in February 2025. Mobile users averaged 3.9 pages per visit in January 2025, ending at 4.06 pages per visit in February.
Dec 24
2.77
–
3.94
–
Jan 25
2.8
+1.35%
3.9
-0.99%
Feb 25
2.92
+4.04%
4.06
+4.11%
The average visit duration for Gemini was 4 minutes and 37 seconds in February 2025
The average visit duration for desktop users was 3 minutes and 36 seconds in January 2025. It slightly increased to 3 minutes and 47 seconds by February. Mobile users had a more extended visit duration. Specifically, it started at 6 minutes and 29 seconds and ended at 6 minutes and 44 seconds in February 2025.
Dec 24
0:03:40
–
0:06:25
–
Jan 25
0:03:36
-1.75%
0:06:29
+1.11%
Feb 25
0:03:47
+4.87%
0:06:44
+3.64%
Google Gemini is no longer a leader in user engagement compared to ChatGPT and Claude
According to the latest Google Gemini statistics, the platform now averages 3.28 pages per visit, which falls behind both ChatGPT (3.81) and Claude (3.79). Similarly, Gemini’s average visit duration is 4 minutes 43 seconds, while users spend more time on Claude (5:53) and significantly longer on ChatGPT (6:47).
Claude also has the lowest bounce rate at 25.92%, compared to 30.94% for ChatGPT and 32.75% for Gemini.
Gemini still shows consistent interaction per user session with 4.2 visits per unique visitor, but this is much lower than ChatGPT’s 11.52 and Claude’s 6.95.
In terms of total traffic, ChatGPT continues to dominate with 3.9 billion monthly visits, followed by Gemini at 284.1 million, and Claude at 73 million.
Monthly visits
284.1M
3.9B
73M
Monthly unique visitors
67.29M
338.4M
10.5M
Visits / Unique visitors
4.2
11.52
6.95
Visit duration
0:04:43
0:06:47
0:05:53
Pages per visit
3.28
3.81
3.79
Bounce rate
32.75%
30.94%
25.92%
Page views
933M
14.87B
276.6M
Google Gemini leads in mobile usage compared to ChatGPT and Claude
Google Gemini statistics show the model’s strong presence in mobile app usage. Specifically, 31.97% of its user base is on mobile devices. In contrast, ChatGPT has a higher desktop usage at 69.59%, while only 30.41% of its users are on mobile. Similarly, Claude has 87.98% of its users on desktop and only 12.02% on mobile.
40% of users utilize Google Gemini for research purposes.
Additionally, 30% of the respondents reported using this language model for creative endeavors. Some use cases include writing poems, scripts, and stories.
Furthermore, 20% indicated that they use Google Gemini for productivity. The examples encompass work or school projects. The remaining 10% of users engage with this language model for entertainment. For instance, they use it to play games and search for videos and music.
Source: SimilarWeb, My Learning
Google Gemini’s traffic distribution shows significant engagement. It spans across various regions, age groups, and marketing channels.
The United States was the top traffic source for Google Gemini in February 2025
As of February 2025, the United States led the traffic sources for Google Gemini with 17.54%. India followed with 9.04%, Indonesia – with 5.25%, Brazil – with 4.48%, and Vietnam – 4.36%. The rest of the world contributed 59.33% of the traffic.
The largest age group of Google Gemini users is 25-34. It makes up 29.7% of the audience
Google Gemini statistics highlight that the 25-34 age group is the most significant demographic. It comprises 29.7% of the audience. The 18-24 age group follows with 21.23%, while the 35-44 group accounts for 19.43%. Smaller segments include the 45-54 group at 14.24%, the 55-64 group at 9.51%, and users aged 65 and above at 5.89%.
Male users represent 57.98% of Google Gemini’s audience
Gender distribution data indicates that male users dominate the platform. This segment made up 57.98% of the audience. Meanwhile, the female counterparts constitute 42.02%.
Direct traffic to Google Gemini peaked at 208.8 million in February 2025
Direct traffic peaked at 208.8 million in February 2025, up from 189.6 million in December 2024. Social media traffic experienced a drop, with a 3.72% decrease from December to February. Organic search traffic grew by 6.04%, while email traffic showed a rapid increase of 20.38%. Similarly, display ads traffic increased by 136% over the same period.
Direct
189.6M
195.2M
208.8M
+10.13%
273.8K
378.6K
329.6K
+20.38%
Referrals
11.05M
11.41M
12.31M
+11.4%
Social
4.57M
4.05M
4.4M
-3.72%
Organic search
47.2M
47.47M
50.05M
+6.04%
Paid search
8.27M
8.22M
7.09M
-14.27%
Display ads
466.1K
905.5K
1.1M
+136%
Source: SimilarWeb
Let’s explore Google Gemini’s cost structure and compare it with other language models.
Gemini 1.5 Flash offers the most cost-effective performance among advanced AI models, with $0.07 (input) and $0.3 (output) per million tokens
Artificial Analysis has carried out research regarding input and output prices. The former implies the cost per token included in the request sent to the API, and the latter – from the API.
Google Gemini statistics indicate that the Gemini 1.5 Flash value is $0.07 per million input or $0.3 per million output tokens. It is a cost-effective option, especially when compared to other models.
For instance, Gemini 1.5 Pro is priced at $1.25 per million input or $5 per million output tokens. On the higher end, GPT-4o costs $2.5 per million input tokens and $10 per million output ones. Claude 3.7 Sonnet is the most expensive, with $3 for input and $15 for output tokens.
Source: Artificial Analysis
Is Google Gemini better than ChatGPT? How does it compare to other language models regarding performance and capabilities? To answer these questions, let’s review Google Gemini statistics below. We’ll focus on the training data, context window, and benchmark performances.
The LaMDA model used in the former version, Google Bard, was trained on the Infinite dataset containing 1.56 trillion words and 137 billion parameters
This massive dataset only required 750 GB of storage. It comprises 12.5% of C4-based data and an equal percentage of code documents from programming tutorials, Q&A websites, and others. Additionally, it includes 6.5% of English web documents and 6.5% of non-English web documents.
Gemini 1.5 Pro boasts a context window of up to two million tokens, the longest of any large-scale foundation model
Gemini 1.5 Pro and 1.5 Flash both have a default context window of up to one million tokens. Thus, these models enable near-perfect recall on long-context retrieval tasks across many formats. The latter include long documents, lines of code, audio, video, and more. For 1.5 Pro, developers and enterprise customers can sign up to try a two-million-token context window.
Comparing the Artificial Analysis Intelligence Index, o3-mini and Grok 3 Reasoning scores the highest at 66, while Gemini 2.0 Flash scores 48
Artificial Analysis has administered the evaluation that incorporated 7 evaluations spanning reasoning, knowledge, math & coding. Let’s review the data in more detail.
o3-mini
66
Grok 3 Reasoning Beta
66
o1
62
DeepSeek R1
60
Claude 3.7 Sonnet Thinking
57
Gemini 2.0 Flash
48
GPT-4o
41
Llama 3.3
41
Mistral Large 2
38
As a result, we can see the comparison of Google Gemini vs GPT-4o. The first leads with a with a score of 48. Meanwhile, GPT-4o scores 41.
Gemini 2.0 Flash is the fastest among its competitors, processing 263 output tokens per second
Google Gemini statistics show that Gemini 2.0 Flash outperforms other models speed-wise. Specifically, it processes 263 tokens per second. Gemini 2.5 Pro follows with 194 tokens per second, and GPT-4o processes 125 tokens per second.
Gemini 2.0 Flash
263
Gemini 2.5 Pro
194
Nova Pro
129
GPT-4o
125
Llama 3.3
121
QwQ-32B
89
Claude 3.7 Sonnet
77
GPT-4o mini
72
DeepSeek V3
33
Among all Gemini models, the 2.5 Pro version has the best performance
Google Gemini statistics by DeepMind prove that the 2.5 Pro version is the top-performing model. According to industry benchmarks, Gemini 2.5 Pro outperform GPT-4.5 at reasoning & knowledge (Humanity’s Last Exam), science (GPQA diamond), mathematics (AIME 2024), and code editing (Aider Polyglot).
Here is detailed table about Google Gemini benchmarks compared to other popular models:
Humanity's Last Exam (Reasoning)
18.8%
6.4%
8.9%
-
8.6%
GPQA diamond (Science)
84.0%
71.4%
78.2%
80.2%
71.5%
AIME 2024 (Mathematics)
92.0%
36.7%
80.0%
93.3%
79.8%
Aider Polyglot (Code editing)
68.6%
44.9%
64.9%
-
56.9%
MMMU (Visual reasoning)
81.7%
74.4%
75.0%
76.0%
-
SimpleQA (Factuality)
52.9%
62.5%
-
43.6%
30.1%
How has Google Bard evolved into Gemini? What key milestones mark this transformation? The timeline below shows the development and enhancement stages of this language model. Let’s take a closer look.
As we can see, Google expands its AI capabilities and reach with each update.
The Google Gemini statistics reveal a robust and versatile language model that excels in various capabilities. With its comprehensive training data and advanced context window, Gemini stands as a competitor to ChatGPT and other leading models.
As seen, Google’s solution offers advanced capabilities and versatile performance across multiple domains. Gemini 2.5 Pro Experimental, for instance, excels in general knowledge, code editing, and mathematical reasoning. Additionally, its context window of up to one million tokens sets it apart from its counterparts.
Google Gemini is poised to influence businesses significantly. Its advanced AI capabilities can enhance various operations, from customer service to data analysis. Thus, companies might be able to drive efficiency and innovation.
Are you looking to develop a custom language model? Don’t hesitate to contact DOIT Software. Our team of experts is ready to help you harness the power of the latest AI technologies. Enhance your operations and engage your customers more effectively today!
Google Gemini statistics reveal there were 67.29 million unique visitors in February 2025 alone.
ChatGPT, Google Bard, and Anthropic Claude are examples of large language models (LLMs). They are designed to generate human-like text based on the input they receive. Thus, they are capable of tasks such as text generation, translation, summarization, etc.
Google rebranded Bard to reflect the significant advancements made to the AI model. In the Google Bard vs Gemini comparison, the latter has improved performance. Moreover, the new version has broader functionalities. This rebranding also aligns with Google’s strategy to continuously innovate and provide state-of-the-art AI solutions.
Google Gemini is up-to-date with information and knowledge available up to early 2025. It might not have information on events that have occurred after that point.