ChatGPT Models used today: different profiles and specific versions available

Every conversation on ChatGPT now begins with GPT-4o, but the user’s concrete experience develops among other technical functions (and limits), automatic model switching, and new opportunities for those who opt for a paid plan.

In mid-2025, accessing ChatGPT means immediately encountering GPT-4o—the model that brings together text, images, and voice for an experience that is increasingly natural and flexible. The platform has chosen to make GPT-4o the default entry point for all new chats so that every user can instantly experience the highest level of multimodal quality available. However, this user experience quickly reveals its boundaries: users without a subscription, once they reach their daily message cap, are automatically shifted to GPT-4.1 mini. This solution ensures continuity, but comes with a reduction in response quality and a narrower range of features, especially on the multimodal front. On the other hand, subscribers can actively select the most suitable model for each context, choosing GPT-4o for creative work, GPT-4.1 and its mini variant for longer sessions or programming, and the o3 and o3-pro series for complex reasoning—adapting every chat to the need of the moment.

Usage statistics confirm the central role of GPT-4o, but show that users’ choices change depending on the type of task and access profile.

The data on model preferences reveals that while users prefer GPT-4o, they turn to specific alternatives when the platform or their work requires it.

Recent analyses of usage habits show that GPT-4o is now at the heart of over 60% of requests handled by the platform. This result comes both from the model’s broad accessibility and from its ability to handle writing, analysis, image generation, and file management tasks with ease. Those who subscribe to Plus or Pro plans keep GPT-4o as their standard, while free users must adapt to an automatic shift to GPT-4.1 mini after reaching their message limit. The adoption of GPT-4.1 and its mini versions is growing steadily, driven by their utility for coding sessions, analysis of long documents, and any activity where service continuity is essential. Meanwhile, models like o3, o3-pro, and o4-mini-high are increasingly chosen by developers, businesses, and advanced users seeking detailed workflow management, custom automations, and explicit reasoning processes. In the background, the role of Custom GPTs is expanding rapidly, now covering more than 12% of daily interactions and pointing to a growing trend toward verticalization and personalization in artificial intelligence.

GPT-4o is the primary model for most users, supporting text, image, and voice (multimodal) with a context window of 128,000 tokens. It is used in over 60% of all ChatGPT web/app conversations. Free users get about 10–15 messages per 3–5 hours before hitting a cap; Plus/Pro/Team users get hundreds of messages per day. It’s ideal for general conversation, creative tasks, file uploads, data analysis, and anything multimodal.

o3 is optimized for advanced reasoning and step-by-step analysis. It is chosen for complex academic, logical, or consulting tasks. Plus/Team users have about 100 messages per week; Pro users get unlimited. o3 currently accounts for roughly 5–10% of advanced user sessions. o3-pro is an enhanced version of o3, focused on long-form and deep reasoning with even greater reliability across lengthy chats. Quotas are the same as o3. It is popular among power-users and professionals for deep-dive technical or research workflows.

o4-mini is a fast and efficient reasoning model best for automations, chatbots, and high-volume tasks. It handles up to 300 messages per day (Plus/Team) and is preferred for situations where speed and throughput matter more than top-level depth. o4-mini-high takes o4-mini further, especially in coding and visual reasoning. With a limit of 100 messages per day (Plus/Team), it’s a top pick for technical troubleshooting and developer support needing fast, consistent responses.

GPT-4.5 (Research Preview) is aimed at creative writing, long content, and idea generation. It has a token window up to 128,000 tokens and is mostly used for stories, scripts, and articles. This model’s availability may vary since it’s in testing.

GPT-4.1 is built for coding, structured analysis, and processing very large documents. It features an extended context window (up to 1 million tokens) and is ideal for users dealing with big datasets, long codebases, or technical reviews. GPT-4.1 mini is a lightweight, fast fallback model—used mostly for quick everyday chats or when users exceed message caps with other models. For Plus/Pro/Team, it’s unlimited; for Free users, it activates automatically after GPT-4o’s cap is reached. While it lacks advanced multimodal and file-handling features, it delivers fast, reliable answers for simple tasks.

The method of model selection varies significantly among free users, paying customers, and organizations, shaping increasingly sophisticated operational strategies.

Users, depending on their profile and context, manage transitions between models through automation, personal preferences, and well-defined internal policies.

For those using the free version, the platform autonomously handles transitions between models: users start with GPT-4o and are moved to GPT-4.1 mini without any direct intervention. Plus and Pro subscribers, however, make flexible use of different models, choosing GPT-4o for creative and brainstorming tasks and turning to GPT-4.1 for working on long documents, coding, or activities that require extended contexts. Within companies, the approach becomes even more layered: teams and advanced professionals often rely on models such as o3-pro and o4-mini-high to solve complex logic problems, manage advanced automations, and oversee step-by-step analyses. Many organizations have adopted policies that reserve GPT-4o for communication and content creation, while GPT-4.1 and its variants are employed for code review and technical documentation. The use of “o” series models is consolidating within research workflows, automation, and data management, while reliance on GPT-3.5 is being gradually phased out and limited mainly to legacy cases via API.

The year 2025 marks the decline of GPT-3.5, the definitive rise of GPT-4o’s multimodality, and the growth of vertical models for increasingly specialized uses.

The real-world distribution of models reveals a profound transformation that reshapes the user experience and shifts the focus toward multimodal and personalized solutions.

Compared to previous years, the true breakthrough in 2025 lies in the final phase-out of GPT-3.5, which now survives only in legacy APIs or in projects where cost is a primary constraint. In daily practice, users now interact almost exclusively with more modern and high-performing models capable of natively integrating text, voice, and visual inputs. GPT-4o thus establishes itself as the reference model, central to creative, multimedia, and productivity-oriented requests. At the same time, the increasing presence of GPT-4.1 and its mini versions responds to the growing demand for long sessions, coding, and large-scale data manipulation, while the “o” series models are gaining traction in automation, advanced analysis, and explicit reasoning, especially in business and professional contexts. The rapid adoption of Custom GPTs confirms the move toward more vertical AI, able to answer the needs of niche and sector-specific tasks.

____________

FOLLOW US FOR MORE.

DATA STUDIOS

datastudios.org