ChatGPT's New Era: GPT-5.3 & GPT-5.4 Unveiled

OpenAI Ushers in a New Era: GPT-5.3 & GPT-5.4 Arrive in ChatGPT

OpenAI is once again redefining the landscape of conversational AI with the rollout of its latest models, GPT-5.3 Instant and GPT-5.4 Thinking and Pro, now arriving for all ChatGPT users. This significant update, effective February 13, 2026, marks a pivotal moment, with several older models like GPT-4o and the previous GPT-5 iterations being retired from ChatGPT. The new generation of AI models promises unprecedented speed, enhanced reasoning, and deeper customization, catering to a broad spectrum of users from everyday learners to enterprise-level professionals.

This strategic evolution aims to consolidate OpenAI's offerings into a more streamlined and powerful user experience. The transition underscores OpenAI's commitment to continuously advance its AI models, ensuring users always have access to the most capable and efficient tools available. As users adopt these new capabilities, the ability of ChatGPT to assist with complex tasks will see a substantial leap forward.

Unpacking the Power of GPT-5.3 Instant & GPT-5.4 Thinking

The core of this update lies in the distinct capabilities of the new GPT-5 models:

GPT-5.3 Instant: Designed as the go-to workhorse for daily tasks and learning, GPT-5.3 Instant excels in delivering quick, clear, and conversational responses. Users will notice marked improvements in handling information-seeking questions, providing detailed how-tos and walk-throughs, generating technical writing, and performing translations. This model maintains a warm, conversational tone, making interactions more engaging and accessible. For more details on its capabilities, you might explore articles on GPT-5.3 Instant.
GPT-5.4 Thinking: This model represents OpenAI's most capable reasoning AI in ChatGPT, specifically engineered for tackling difficult, real-world problems. Compared to its predecessors, GPT-5.4 Thinking demonstrates superior prowess in areas such as spreadsheet creation and editing, producing polished frontend code, generating slideshows, solving hard math problems, comprehensive document and image understanding, advanced instruction following, tool use, and sophisticated research tasks that require synthesizing information from diverse web sources. It can 'think' longer on complex challenges without timing out, effectively tracking its progress and reducing the need for users to re-state important details. A key feature is its ability to display a short preamble before reasoning commences, allowing users to provide additional instructions to adjust its direction before the final answer is complete. Outputs are also more streamlined, featuring cleaner formatting and less unnecessary header text.
GPT-5.4 Pro: For the most demanding tasks and extensive, long-running workflows, GPT-5.4 Pro offers the highest capability option within the GPT-5.4 family in ChatGPT. This model is tailored for research-grade intelligence, providing unparalleled performance for professional and enterprise-level applications that require the utmost precision and depth.

Navigating ChatGPT's Advanced Model Selection and Reasoning

With the introduction of these new language models, ChatGPT offers sophisticated mechanisms for model selection to optimize user experience:

Auto Selection: When 'Auto' is selected in ChatGPT, the system intelligently determines whether to use GPT-5.3 Instant for quick responses or GPT-5.4 Thinking for more complex tasks requiring deeper reasoning. This dynamic switching ensures users always get the most appropriate model for their query without manual intervention.
Manual Selection: Users can also directly choose their preferred model from the model picker: 'Auto', 'Instant' (GPT-5.3 Instant), 'Thinking' (GPT-5.4 Thinking), or 'Pro' (GPT-5.4 Pro).
Thinking Traces and Preamble: When GPT-5.4 Thinking or GPT-5.4 Pro is engaged, especially for intricate problems, it may begin with a brief preamble outlining its planned approach. Users can intervene during this 'thinking time' to steer the model's direction, offering real-time guidance. While manual selection of 'Thinking' always displays a trace, 'Auto' mode may only show it for longer reasoning sequences.
Thinking Time Toggle: For Plus and Business users, a 'thinking-time' toggle in the message composer provides control over the depth of reasoning. Options like 'Standard' (balancing speed and intelligence) and 'Extended' (deeper processing) are available. Pro users gain 'Light' for the snappiest responses and 'Heavy' for the most profound reasoning, allowing fine-tuning for specific task requirements.

Understanding Tiered Access, Usage Limits, and Context Windows

Access to the new GPT-5 models varies across ChatGPT tiers, accompanied by specific usage limits and enhanced context windows designed to support diverse user needs.

GPT-5.3 Instant is broadly available across all ChatGPT tiers. However, access to the more advanced GPT-5.4 Thinking and Pro models, along with manual model selection capabilities, is tier-dependent. Enterprise and Edu workspaces have GPT-5.3 Instant and GPT-5.4 Thinking/Pro default off, requiring admin enablement.

ChatGPT Model Availability, Usage Limits, and Context Windows

ChatGPT Tier	GPT-5.3 Instant Usage Limits	GPT-5.3 Instant Context Window	GPT-5.4 Thinking Usage Limits	GPT-5.4 Thinking Context Window	GPT-5.4 Pro Access
Free	10 messages / 5 hours	16K	N/A	N/A	N/A
Plus/Go	160 messages / 3 hours	32K	3,000 messages / week (Plus); 10 msg / 5 hrs (Go)	256K (128K input + 128K output)	N/A
Business	Unlimited (subject to guardrails)	32K	Unlimited (subject to guardrails)	256K (128K input + 128K output)	Yes
Pro	Unlimited (subject to guardrails)	128K	Unlimited (subject to guardrails)	400K (272K input + 128K output)	Yes
Enterprise/Edu	Unlimited (subject to guardrails)	128K	Unlimited (subject to guardrails)	256K (128K input + 128K output)	Yes

Note: Unlimited access is subject to OpenAI's Terms of Use and abuse guardrails. Automatic switching from GPT-5.3 Instant to GPT-5.4 Thinking does not count towards manual selection limits.

The expanded context window sizes represent a significant advancement, allowing the models to process and remember vastly more information within a single conversation. This capability is particularly beneficial for users engaged in long-form writing, complex data analysis, or multi-step problem-solving, as it reduces the need for repetitive prompting and ensures greater conversational coherence. For business users, this contributes significantly to scaling AI for everyone.

Enhanced Tool Integration and Future-Proofing with GPT-5 Models

The new GPT-5.3 Instant and GPT-5.4 Thinking models seamlessly integrate with all existing tools within ChatGPT, ensuring a versatile and powerful user experience. This includes essential features such as:

Web search: For up-to-date information retrieval.
Data analysis: To process and interpret complex datasets.
Image analysis & Image generation: Expanding creative and analytical capabilities.
File analysis: For deeper understanding of uploaded documents.
Canvas: For visual interactions and content creation.
Memory: Allowing the AI to retain context across sessions.
Custom Instructions: For tailored AI behavior and responses.

This comprehensive tool support means that users can leverage the full spectrum of ChatGPT functionalities with the enhanced intelligence of the GPT-5 models, streamlining workflows and expanding possibilities across various tasks.

Personalizing Your ChatGPT Experience: Tone, Style & Legacy Access

OpenAI has significantly upgraded ChatGPT's AI updates with enhanced controls for tone and style, allowing users to personalize their interactions more deeply than ever before. New refined personality presets are available, including 'Default,' 'Friendly' (formerly Listener), 'Efficient' (formerly Robot), 'Professional,' 'Candid,' and 'Quirky.' For those seeking even more distinct characteristics, 'Cynical' and 'Nerdy' options are available under Personalization settings.

Beyond personality presets, users can fine-tune response conciseness, warmth, scannability, and emoji frequency. These personalization options apply instantly across all chats, continuously improving over time as the rollout progresses. Crucially, GPT-5.3 and GPT-5.4 exhibit improved adherence to Custom Instructions, granting users precise control over the AI's behavior and conversational style, making ChatGPT a truly adaptive and personal AI assistant.

As part of this rollout, several older models, including GPT-4o, GPT-4.1, GPT-4.1 mini, and the original GPT-5 (Instant and Thinking), were retired from ChatGPT as of February 13, 2026. While API access remains unchanged for these legacy models, within the ChatGPT interface, conversations created with previous models will now run on the new GPT-5.3 and GPT-5.4 equivalents. GPT-5.2 Thinking will remain available in 'Legacy Models' for 90 days post-GPT-5.4 Thinking launch, specifically for Plus and Pro users, easing the transition to these powerful new AI models.

Original source

https://help.openai.com/en/articles/11909943-gpt-53-and-52-in-chatgpt

Frequently Asked Questions

What are the key differences between GPT-5.3 Instant, GPT-5.4 Thinking, and GPT-5.4 Pro in ChatGPT?

GPT-5.3 Instant is engineered as a rapid and efficient workhorse for daily tasks, excelling in information retrieval, how-to guides, technical writing, and translation, while maintaining a friendly, conversational tone. GPT-5.4 Thinking represents OpenAI's most advanced reasoning model for challenging real-world problems, demonstrating superior performance in tasks like spreadsheet creation, complex coding, slideshow generation, advanced mathematics, document and image understanding, and multi-source research. It can 'think' longer and track progress effectively, often showing a preamble. GPT-5.4 Pro is the highest-capability variant of GPT-5.4, designed for the most demanding tasks and extensive, long-running workflows, providing research-grade intelligence for professional and enterprise-level applications. Each model caters to different user needs, balancing speed, depth of reasoning, and task complexity.

How does the 'Auto' model selection work in ChatGPT, and when should I use it versus manual selection?

The 'Auto' feature in ChatGPT intelligently switches between GPT-5.3 Instant and GPT-5.4 Thinking based on the complexity of your request. For simpler, everyday queries, it defaults to GPT-5.3 Instant for speed. If a task requires deeper reasoning, such as complex problem-solving or multi-step analysis, 'Auto' will automatically engage GPT-5.4 Thinking. Manual selection is beneficial when you have a clear understanding of your task's requirements. For instance, if you know you need quick, concise answers, you'd manually select 'Instant.' If you're tackling a highly complex research task that demands thorough analysis and multi-step reasoning, directly choosing 'Thinking' or 'Pro' ensures the model allocates sufficient resources from the start, bypassing the 'Auto' detection process and potentially saving time on intricate workflows. It also ensures you get the thinking trace even for short reasoning sequences.

What are the usage limits for the new GPT-5.3 and GPT-5.4 models across different ChatGPT tiers?

Usage limits for GPT-5.3 and GPT-5.4 vary significantly by ChatGPT tier. Free users can send up to 10 messages with GPT-5.3 every 5 hours before switching to a mini version. Plus/Go users have a higher limit of 160 GPT-5.3 messages every 3 hours. For GPT-5.4 Thinking, Plus and Business users can manually select it for up to 3,000 messages per week, after which it becomes unselectable (though 'Auto' can still route to it). Go users can access Thinking for 10 messages every 5 hours. ChatGPT Business and Pro plans generally offer unlimited access to GPT-5 models, subject to abuse guardrails and Terms of Use. It's crucial to be aware of these limits to optimize your usage and avoid interruptions, especially for critical tasks requiring advanced reasoning.

Can I customize the personality and tone of ChatGPT with the new GPT-5 models?

Yes, ChatGPT with GPT-5.3 and GPT-5.4 models introduces enhanced customization options for tone and style. Users can now choose from refined personality presets like Default, Friendly, Efficient, Professional, Candid, and Quirky. Additionally, under Personalization settings, 'Cynical' and 'Nerdy' are available. These controls also allow users to adjust response conciseness, warmth, scannability, and emoji frequency. These personalization options are applied across all chats instantly and are being rolled out gradually. Crucially, the new GPT-5 models demonstrate improved adherence to Custom Instructions, giving users even more precise control over the AI's behavior and conversational style, making interactions more tailored to individual or professional needs.

What is the 'thinking time' toggle in GPT-5.4 Thinking, and how does it affect responses?

The 'thinking time' toggle in GPT-5.4 Thinking, available in the message composer for Plus and Business users, allows you to control the depth and speed of the model's reasoning. 'Standard' is the new default, balancing speed and intelligence. 'Extended' offers deeper processing, previously the default for Plus. Pro users gain two additional options: 'Light' for the snappiest responses and 'Heavy' for the most profound reasoning. Selecting 'Light' or 'Standard' results in quicker answers that still apply reasoning, ideal for balanced speed and intelligence. 'Extended' or 'Heavy' provides more time for deeper, comprehensive responses, suitable for high-stakes or difficult questions. This feature helps tailor the AI's processing to the immediate task, optimizing for either speed or thoroughness, though it's currently only available on the ChatGPT Web interface.

Are legacy GPT models still accessible in ChatGPT after the GPT-5.3 and GPT-5.4 rollout?

As of February 13, 2026, several older models including GPT-4o, GPT-4.1, GPT-4.1 mini, OpenAI o4-mini, and GPT-5 (Instant and Thinking) have been retired from ChatGPT, meaning they are no longer directly available for selection. API access for these models remains unchanged, but within the ChatGPT interface, they have been superseded by the new GPT-5.3 and GPT-5.4 equivalents. Notably, GPT-5.2 Thinking will remain available in 'Legacy Models' for a transitional period of 90 days after the launch of GPT-5.4 Thinking, specifically for Plus and Pro users. Older conversations that were generated with previous models will now run on the new GPT-5.3 and GPT-5.4 engines, which means their outputs might differ if you continue those chats.

How do the new context windows for GPT-5 models enhance user experience in ChatGPT?

The significantly expanded context windows for GPT-5.3 Instant and GPT-5.4 Thinking models in ChatGPT dramatically improve the AI's ability to maintain coherence and understand complex, lengthy conversations. For instance, GPT-5.3 Instant offers up to 128K tokens for Pro/Enterprise users, while GPT-5.4 Thinking boasts up to 400K tokens for Pro users. This means the models can process and remember a much larger amount of information within a single interaction, reducing the need for users to repeat details and allowing for more nuanced, extended dialogues. This enhanced 'memory' is particularly beneficial for deep research tasks, extensive code reviews, or detailed document analysis, enabling the AI to grasp intricate details and provide more relevant, context-aware responses over prolonged exchanges.

Stay Updated

Get the latest AI news delivered to your inbox.