ChatGPT, OpenAI’s text-generating AI chatbot, has taken the world by storm since its launch in November 2022. In 2025, the company battled perceptions of falling behind rivals like DeepSeek, while ramping up data center projects, securing billions in funding, and focusing on making ChatGPT more capable and safer. The year ended with over 800 million weekly active users and a host of new features. Below is a detailed timeline of the most significant updates throughout 2025.

January 2025

OpenAI launched o3-mini, its latest reasoning model, positioning it as both powerful and affordable. The company also introduced ChatGPT Gov for U.S. government agencies, offering enhanced security and compliance. A new beta feature called tasks allowed users to set reminders and recurring actions. Additionally, OpenAI began testing signups using only a phone number, and a survey showed teens increasingly using ChatGPT for schoolwork, with 26% of U.S. teens aged 13-17 reporting usage.

February 2025

OpenAI canceled the standalone o3 model in favor of a unified GPT-5 release. It launched deep research, an AI agent for complex multi-source research. ChatGPT web search was made available without login. The company also revealed more of the o3-mini model’s reasoning chain of thought. An analysis by Epoch AI found that ChatGPT queries actually consume far less energy than previously assumed—about 0.3 watt-hours per query with GPT-4o.

March 2025

OpenAI announced plans to release an open language model for the first time since GPT-2. It upgraded ChatGPT’s image generation capabilities using GPT-4o, which went viral for creating Studio Ghibli-style images, raising copyright concerns. The company introduced new tools for building AI agents via the Responses API, and reportedly planned to charge up to $20,000 per month for specialized agents like “PhD-level research.” ChatGPT weekly active users doubled to 400 million by February, driven by new releases.

April 2025

OpenAI launched o3 and o4-mini reasoning models, which could use web browsing, coding, and image processing—but hallucinated more than previous models. It also released GPT-4.1 focusing on coding (available via API, not ChatGPT). The company offered ChatGPT Plus free to U.S. and Canadian college students through May. Over 700 million images were generated in ChatGPT since the upgraded image generator was released on March 25. OpenAI also began testing a new feature that remembers previous conversations to personalize responses.

May 2025

The company introduced its AI coding agent Codex, powered by codex-1. Sam Altman expressed desire for ChatGPT to track every aspect of a person’s life to become more personalized. GPT-4.1 models were made available directly in ChatGPT for coding tasks. Deep research gained GitHub connectivity in beta. OpenAI also announced a data residency program in Asia and the “OpenAI for Countries” initiative to expand AI infrastructure globally. After a sycophancy issue, OpenAI promised to prevent overly agreeable responses in future updates.

June 2025

OpenAI launched o3-pro, an enhanced reasoning model for Pro and Team users. Advanced Voice Mode was upgraded for all paid users, offering more natural conversations. New enterprise features included meeting recording, integrations with Google Drive and Box, and MCP support. A study from MIT suggested ChatGPT may harm critical thinking skills by reducing neural engagement during writing tasks. The iOS app was downloaded 29.6 million times in 28 days, rivaling social media apps. Altman also clarified that the energy for a typical query is minimal.

July 2025

OpenAI introduced Study Mode in ChatGPT to promote critical thinking in students. It launched ChatGPT Agent, a general-purpose agent capable of automating computer-based tasks like calendar management, code running, and shopping. ChatGPT reached 2.5 billion daily prompts. Altman warned users against treating ChatGPT as a therapist due to lack of confidentiality. Stanford researchers flagged risks with therapy-style chatbots. OpenAI delayed its open model release for additional safety testing, and reports emerged of an upcoming AI browser.

August 2025

OpenAI unveiled GPT-5, a smarter, task-ready model that automatically chooses between Fast and Thinking modes. The company offered ChatGPT Enterprise to federal agencies for just $1 for the first year. It also returned to open source with two new open-weight models: gpt-oss-120b and gpt-oss-20b. ChatGPT neared 700 million weekly active users, quadrupling in a year. The mobile app hit $2 billion in revenue. xAI filed a lawsuit alleging Apple and OpenAI colluded. OpenAI also tightened ChatGPT safeguards after a teen suicide lawsuit.

September 2025

Parental controls were rolled out to limit sensitive content and set quiet hours for teens. ChatGPT Pulse was introduced, delivering personalized morning briefs to Pro users. OpenAI launched Instant Checkout, allowing U.S. users to purchase products from Etsy and Shopify directly within ChatGPT. The budget-friendly ChatGPT Go plan expanded to Indonesia. GPT-5-Codex was released, capable of tackling coding tasks from seconds to seven hours. The Model Behavior team was restructured, with founding leader Joanne Jang spinning up OAI Labs for prototyping human-AI collaboration.

October 2025

OpenAI revealed that ChatGPT handles over a million suicide-related conversations weekly and improved responses with mental health expert input. The company launched ChatGPT Atlas, an AI browser starting on Mac, aiming to replace traditional search. It partnered with Walmart for shopping features and expanded the ChatGPT Go plan to 16 Asian countries. ChatGPT surpassed 800 million weekly active users. Developers were allowed to build interactive apps directly inside ChatGPT, with partners like Booking.com, Spotify, and Figma. App growth slowed but still saw millions of daily users.

November 2025

OpenAI introduced an AI shopping assistant for holiday purchases and integrated voice mode directly into the main chat interface. Group chats became available to all users. GPT-5.1 was released with Instant and Thinking variants, offering warmer tone and better reasoning. A Munich court ruled ChatGPT violated music copyright on nine songs. OpenAI explored consumer health tools and reached 1 million business clients. Seven more families sued OpenAI over ChatGPT-related suicides, alleging premature release of GPT-4o without safeguards.

December 2025

OpenAI added controls to tweak ChatGPT’s warmth, enthusiasm, and emoji use. It updated guidelines for teen safety and hit $3 billion in mobile app revenue faster than TikTok. GPT Image 1.5 was released with improved instruction following. Disney invested $1 billion in OpenAI, giving one-year exclusive use of its characters in Sora. OpenAI reported enterprise message volume up 8x since late 2024. The company unveiled GPT-5.2 in three versions: Instant, Thinking, and Pro. CEO Sam Altman declared a “code red” to prioritize ChatGPT improvement amid rising competition from Google.

Source: TechCrunch News