ChatGPT-5 Statistics By User, API Pricing, Subscription Plans And Trends (2025)

Introduction

ChatGPT-5 Statistics: GPT-5 is known to be the most advanced AI system so far. GPT-5, which is smarter than any other earlier ChatGPT models and excels in many areas like coding, math, writing, health, and even understanding images.

This new AI knows when to give quick answers and when to take more time to think deeply, delivering expert-level responses every time. Everyone can use GPT-5, but Plus subscribers get more access, while Pro subscribers enjoy GPT-5 Pro—a special version with extra reasoning power for even better and more accurate answers.

Editor’s Choice

• As mentioned in lifearchitect.ai, the organisation behind GPT-5 is OpenAI.
• ChatGPT-5 was publicly released on August 7, 2025.
• Gadgets360 clearly stated that there are four different versions: the already known GPT-5 base model, GPT-5 Mini, GPT-5 Nano, and a new version called GPT-5 Chat.
• As mentioned in help.openai.com, users on the free ChatGPT-5 plan can send up to 10 messages every 5 hours. When they hit this limit, their chats will automatically switch to the mini version of the model until the limit resets.
• According to Hindustan Times, the GPT-5 API costs USD 1.25 for every million input tokens and USD 10 for every million output tokens.
• OpenAI offers ChatGPT 5 in five different subscription plans: Free, Plus, Pro, Team, and Enterprise.
• hindustantimes.com also shows that it supports a massive 256,000-token context window, up from 200,000 tokens in the o3 model, enabling better handling of long conversations, code, and documents.
• As mentioned in Business Insider, when GPT-5 was released, it became available to all 700 million people who were using ChatGPT at that time.
• In coding benchmarks, GPT-5 scored 74.9% on SWE-Bench Verified, 55% on SWE-Lancer, and 88% on Aider Polyglot, showing improved multi-language support and bug fixing.
• HealthBench Hard results showed GPT-5 thinking at 25.5%, compared to O3’s 31.6%, but with higher-quality medically reviewed outputs.
• Hallucination rates dropped by 65% versus o3 and 26% versus GPT-4o, reflecting safer and more accurate responses.
• OpenAI reports over 5,000 hours of red teaming and testing to reduce unsafe outputs and manage dual-use prompts with “safe completions.”

About ChatGPT-5

• As mentioned in lifearchitect.ai, the organisation behind GPT-5 is OpenAI.
• The model is known internally by the project names Gobi and Arrakis.
• GPT-5 is a multimodal model, meaning it can understand and work with different types of data.
• The training data used for GPT-5 included approximately 114 trillion tokens.
• Training began in January 2025 and was completed by April 2025, with the data used for training going up until September 2024.
• GPT-5 was publicly released on August 7, 2025.
• The latest GPT-5 model, with an estimated 300 billion parameters, finished training in April 2025 using 989 H100 chips for 50 days, which is roughly 16,439 years of chip time and cost USD 612 million.
• Its MMLU score is not listed here. Other models like Gemini and Grok 3 show similarly large-scale training efforts with costs reaching into the hundreds of millions or over a billion dollars.

Features of ChatGPT-5

• ChatGPT 5 offers better reasoning skills and gives more precise answers.
• It can understand pictures, talk using voice, and respond in a way that feels more caring and thoughtful.
• The voice feature, available for free and paid users, will have higher usage limits than GPT-4.
• It will also be added to custom GPTs, and the Standard Voice Mode will be removed in 30 days.
• When memory is turned on, it remembers what users like, making conversations over time feel more personal and natural.
• Users just need to type their prompt, and GPT-5 will handle all the hard work.
• Additionally, OpenAI has improved GPT-5 to reduce the problem of making things up, making it more accurate than the previous version.
• ChatGPT 5 will offer many ways to customise it, giving users a much more personal experience.

OpenAI GPT Family Analysis (2018 to 2026)

(Source: pcdn.co)

Working Process of ChatGPT-5

• When someone chooses GPT-5 in ChatGPT, the system automatically picks whether to respond in Chat mode or Thinking mode.
• For harder questions, GPT-5 changes to Thinking mode, where it takes more time to think carefully before answering.
• This choice to “think longer” is based on clues from the person’s question and conversation, as well as patterns the system learned from how people usually pick models, their preferences, and how often the model’s answers are right.

ChatGPT-5 User Statistics

• As mentioned in Business Insider, when GPT-5 was released, it became available to all 700 million people who were using ChatGPT at that time.
• The number of ChatGPT users grew a lot, reaching 500 million weekly users by March 2025.
• This is four times more than the number of users a year earlier.
• Windows Central reported that people use ChatGPT by sending around 2.5 to 3 billion prompts every day.
• On average, they spend about 16 minutes chatting with it each day.
• New user features include four selectable personalities (Cynic, Robot, Listener, Nerd) and customizable chat colors, with Advanced Voice Mode integration planned, as per the reports of Hindustan Times.

Usage Limits of ChatGPT 5 Statistics

• As mentioned in help.openai.com, users on the free ChatGPT plan can send up to 10 messages every 5 hours. When they hit this limit, their chats will automatically switch to the mini version of the model until the limit resets. Free users also get access to one GPT-5 Thinking message each day.
• People with a ChatGPT Plus subscription can send up to 80 messages every 3 hours. After reaching this limit, their chats will also change to the mini version of the model until the limit resets.
• Those on Plus or Team plans can manually choose the GPT-5 Thinking model from the model selector, but this option has a weekly limit of 200 messages. When they use all 200 messages in a week, a notification will pop up, and the GPT-5 Thinking model will no longer be available to select until next week.

GPT-5 API Pricing Statistics

• According to Hindustan Times, the GPT-5 API costs USD 1.25 for every million input tokens and USD 10 for every million output tokens.
• For the smaller GPT-5-mini version, the price is USD 0.25 per million input tokens and USD 2 per million output tokens.
• The cheapest option, GPT-5-nano, is priced at USD 0.05 per million input tokens and USD 0.40 per million output tokens.
• Notably, the nano model is now cheaper than Google’s Gemini 2.5 Flash.

ChatGPT-5 Subscription Plans Statistics

GPT-5 Benchmarks Statistics by Coding Performance

(Source: datacamp.com)

• On the SWE-Bench Verified test, which measures real-world Python coding skills, GPT-5 achieves a score of 74.9%.
• This is higher than OpenAI’s O3 model, which scored 69.1%, and much better than GPT-4.1, which scored 54.6%.
• When working with difficult problems, GPT-5 is more efficient: it uses 22% fewer output tokens and makes 45% fewer calls to external tools compared to the o3 model to reach its results.
• In the Aider Polyglot test, which evaluates coding in multiple programming languages, GPT-5 scores 88%.
• This improves on the O3 model’s 81%, meaning GPT-5 reduces errors by about one-third.

By Math and Scientific Reasoning, Statistics

• A report published by DataCamp shows that GPT-5 performs very well on tough math tests, such as the 2025 AIME math competition, where it scored 94.6%, which is higher than the O3 model’s 88.9%.
• In the Harvard-MIT Math Tournament (HMMT), GPT-5 scored 93.3% without using any tools, beating O3’s score of 85%.
• On FrontierMath, which is a difficult expert-level math test that lets the model use a Python tool, GPT-5 scored 26.3%.
• Although this number isn’t very high, it’s still better than the O3 model’s 15.8%.
• For PhD-level science questions in the GPQA Diamond test, GPT-5 scored 87.3% when using tools like Python, and 85.7% without tools.

Multimodal Reasoning Statistics

• GPT-5 achieved a new record on multimodal tests.
• com also states that it scored 84.2% on MMMU, which measures college-level visual reasoning, better than the previous O3 model.
• On MMMU-Pro, a harder test at the graduate level, GPT-5 scored 78.4%, again beating o3.
• For video reasoning with up to 256 frames, called VideoMMMU, GPT-5 reached 84.6% accuracy, higher than O3’s 83.3%.
• When interpreting scientific figures (CharXiv Reasoning), GPT-5 scored 81.1% with its thinking mode on, outperforming O3.
• In spatial reasoning tasks (ERQA), GPT-5 achieved 65.7%, also ahead of O3.

By Humanity’s Last Exam (HLE) Statistics

• Humanity’s Last Exam is a tough test made up of 2,500 carefully chosen PhD-level questions. It covers subjects like math, physics, chemistry, linguistics, and engineering.
• According to OpenAI’s official results, GPT-5 scores 24.8% when it works without any extra tools.
  When using its Pro version with tools, GPT-5’s score improves to 42.0%.
• Grok 4, which is from xAI, gets about 26% without tools and 41.0% with tools.
• Grok 4 Heavy uses several agents working together and combines their answers with a higher score of 50.7%.

Recent Activity Statistics by ChatGPT-5

• As of August 6, 2025, the total number of tokens processed per day was 1.2 billion, of which prompt tokens accounted for 1.15 billion and completion tokens (50.7 million).
• Meanwhile, on August 7, the number increased, resulting in prompt tokens (3.19 billion) and completion tokens (1.21 million).

By ChatGPT-5 Mini (2025)

(Source: openrouter.ai)

By ChatGPT-5 Nano (2025)

(Source: openrouter.ai)

Apps Using Statistics of ChatGPT-5

• According to reports from Openrouter.ai, among well-known AI coding and chatbot services, Cline is at the top, processing 834 million tokens using ChatGPT-5.
• It provides an independent coding assistant that works right inside your IDE.

Furthermore, other apps using GPT-5 Chat are stated in the table below:

You may also be interested in:

• Generative Ai Statistics
• Perplexity Ai Statistics
• Grammarly Ai Statistics
• Meta Ai Statistics
• Ai In Business Statistics

Conclusion

GPT-5 marks a major step forward in artificial intelligence. It understands things better, can remember much more information, and is much stronger at coding. The model is safer to use and makes fewer mistakes, so talking to it feels more trustworthy.

Users also get new choices like picking different personalities and voice options, making the experience more personal. As many people start using GPT-5 and it keeps improving, it is changing the way we interact with AI. This opens up many new opportunities for work and daily life across different fields.

Maitrayee Dey

Maitrayee Dey has a background in Electrical Engineering and has worked in various technical roles before transitioning to writing. Specializing in technology and Artificial Intelligence, she has served as an Academic Research Analyst and Freelance Writer, particularly focusing on education and healthcare in Australia. Maitrayee's lifelong passions for writing and painting led her to pursue a full-time writing career. She is also the creator of a cooking YouTube channel, where she shares her culinary adventures.At Smartphone Thoughts, Maitrayee brings her expertise in technology to provide in-depth smartphone reviews and app-related statistics, making complex topics easy to understand for all readers.

More Posts By Maitrayee Dey

ChatGPT-5 Statistics By User, API Pricing, Subscription Plans And Trends (2025)

How Many Videos Are On YouTube? Statistics And Facts, By Country, Demographics, Trends (2025)

How Many Users Are On X (Formerly Twitter)? Key Statistics and Facts (2025)

How Many Users On TikTok? Statistics And Trends (2025)

How Many Users Does Reddit Have? (2025)

How Many SEO Keywords Should I Use Per Page?

How Many People Work At X? (2025)

How Many People Work At Tesla? Statistics And Trends (2025)

How Many People Work At Midjourney? Statistics By Region And Category (2025)

How Many People Work at Nvidia? Statistics And Facts (2025)