ChatGPT 5.5 just launched: features, performance, benchmarks, limits, and more
- 11 minutes ago
- 9 min read
GPT-5.5 is now live inside ChatGPT, and this release is being presented by OpenAI as a meaningful step forward in reasoning quality, agentic work, coding, research, and document-heavy execution rather than as a routine silent refresh.
Inside ChatGPT, the user-facing layer is GPT-5.5 Thinking, while higher-end plans also get GPT-5.5 Pro, which OpenAI describes as the stronger premium layer for harder questions and higher-accuracy work.
The launch is already live in ChatGPT and Codex, while the API is still described by OpenAI as coming very soon, which means the first real impact of GPT-5.5 is happening at product level before the full developer rollout arrives.
What changes most here is the overall shape of difficult work inside ChatGPT, because OpenAI is claiming stronger performance in agentic coding, computer use, knowledge work, data analysis, information synthesis, and long multi-step tasks, while also arguing that GPT-5.5 reaches that higher level without becoming slower than GPT-5.4 in real-world serving.
The practical result is a release that affects more than raw model quality, since it also reshapes plan access, model tiers, legacy-model handling, and the way premium users will think about difficult work inside ChatGPT from now on.
··········
GPT-5.5 is now live in ChatGPT, and OpenAI is treating it as a major release.
The launch is already active inside ChatGPT, with GPT-5.5 Thinking as the main reasoning layer and GPT-5.5 Pro as the higher-end premium layer.
OpenAI’s launch page says GPT-5.5 was announced on April 23, 2026, and that it is rolling out in ChatGPT and Codex now, while the API is still pending.
OpenAI’s Help Center separately states that “GPT-5.5 Thinking is our most capable reasoning model in ChatGPT,” which makes clear that the practical in-product identity of the launch is tied to the Thinking layer rather than to the plain model name alone.
The same launch also introduces GPT-5.5 Pro inside ChatGPT for higher-end tiers, which gives the product a clearer internal premium split than before.
........
· GPT-5.5 is officially launched.
· In ChatGPT, the main user-facing version is GPT-5.5 Thinking.
· GPT-5.5 Pro is a separate higher-end layer.
........
The launch in one view
Area | Current official position |
Model | GPT-5.5 |
ChatGPT-facing version | GPT-5.5 Thinking |
Higher-end version | GPT-5.5 Pro |
Live now in | ChatGPT, Codex |
API status | Coming very soon |
··········
The release changes ChatGPT more through reasoning quality and workflow depth than through a simple version jump.
OpenAI is positioning GPT-5.5 as a stronger model for agentic work, coding, research, data analysis, and document-heavy tasks.
OpenAI describes GPT-5.5 as its smartest and most intuitive model yet, and the company’s official framing puts the emphasis on practical work that involves ambiguity, tool use, long context, and multi-step execution rather than on a narrow benchmark headline alone.
The launch materials say GPT-5.5 is especially strong in agentic coding, computer use, knowledge work, and early scientific research, while also stressing that the model can carry more of the task forward with less user micromanagement.
That positioning matters because it moves the center of gravity away from simple “smarter answers” language and toward a product story built around harder tasks, longer sessions, better continuation, and stronger execution through tools.
··········
GPT-5.5 Thinking and GPT-5.5 Pro now create a clearer two-layer premium structure inside ChatGPT.
The launch separates a broadly available advanced reasoning layer from a more premium layer meant for higher-accuracy and higher-difficulty work.
OpenAI’s official material makes the split explicit.
GPT-5.5 Thinking is the main advanced reasoning layer that reaches a wider set of ChatGPT plans, while GPT-5.5 Pro is the version OpenAI says is designed for harder questions and higher-accuracy output.
That product structure changes how the launch should be understood, because ChatGPT is no longer simply getting one new model and is instead getting a more visible internal tiering of reasoning capability.
This matters most for users doing very demanding work, since the Pro layer is being positioned as the place where OpenAI pushes its strongest accuracy and hardest-task handling.
........
· GPT-5.5 Thinking is the broader advanced reasoning layer.
· GPT-5.5 Pro is the more premium and higher-accuracy layer.
· ChatGPT now has a clearer internal capability split.
........
How the new ChatGPT model layers are positioned
Layer | Current official role |
GPT-5.5 Thinking | Main advanced reasoning layer in ChatGPT |
GPT-5.5 Pro | Harder questions, higher-accuracy premium layer |
··········
Availability depends heavily on plan, and that matters almost as much as the launch itself.
GPT-5.5 is not distributed equally across all ChatGPT tiers, and the product logic changes substantially by plan.
OpenAI says GPT-5.5 Thinking is rolling out to Plus, Pro, Business, and Enterprise users in ChatGPT, while GPT-5.5 Pro is rolling out to Pro, Business, and Enterprise users.
The Help Center also says Go users can access Thinking through the tools menu, although that route comes with a much smaller cap and should be understood as a narrower access path rather than as full parity with the broader paid tiers.
This plan gating is a central part of the launch because the practical meaning of GPT-5.5 depends not only on the model’s quality, but also on whether the user receives the broader Thinking layer, the Pro layer, or a more limited path through the product.
........
· GPT-5.5 Thinking reaches several paid ChatGPT tiers.
· GPT-5.5 Pro is reserved for higher-end tiers.
· Go gets a more limited access path.
........
Who gets what in ChatGPT
Plan | GPT-5.5 Thinking | GPT-5.5 Pro |
Go | Limited tools-menu access | No |
Plus | Yes | No |
Pro | Yes | Yes |
Business | Yes | Yes |
Enterprise | Yes | Yes |
··········
Usage limits are one of the most important practical parts of the release.
The value of GPT-5.5 depends not only on capability, but also on how far each plan can actually use it before hitting the ceiling.
OpenAI’s Help Center says Go users get 10 GPT-5.5 Thinking messages every 5 hours after enabling it through tools, while Plus and Business users can manually select GPT-5.5 Thinking with a limit of up to 3,000 messages per week.
OpenAI also describes Pro and Business as offering much broader GPT-5 access, with pricing-page language around unlimited GPT-5.5 messages or more generous use subject to abuse guardrails and product policy.
Those ceilings are not a side detail, because the usefulness of a premium reasoning model changes drastically once the user is doing repeated coding sessions, long research workflows, or document-heavy analysis across the same week.
........
· Plus and Business get much broader GPT-5.5 Thinking access than Go.
· Go has a much tighter cap.
· Pro and Business sit closer to the fullest GPT-5.5 experience.
........
GPT-5.5 limits in current official materials
Plan / layer | Current official limit signal |
Go | 10 GPT-5.5 Thinking messages every 5 hours |
Plus / Business manual Thinking selection | Up to 3,000 messages per week |
Pro / Business broader GPT-5 access | Much broader, subject to guardrails |
··········
Performance is where OpenAI is trying to show that GPT-5.5 is more than a routine iteration.
The company is framing GPT-5.5 as stronger than GPT-5.4 in agentic coding, computer use, knowledge work, and long multi-step execution.
OpenAI’s official launch claims present GPT-5.5 as a meaningful step over GPT-5.4 in the kinds of work that matter most to advanced ChatGPT users, especially when those tasks involve tool use, planning, ambiguity, and execution across longer chains of reasoning.
The company is also tying the model to research, data analysis, information synthesis, and document-heavy work, which broadens the performance story beyond coding and makes the release relevant to a much wider professional audience.
This matters because the product claim is no longer “slightly better answers,” and is instead “more capable handling of difficult work inside ChatGPT,” which is a much stronger and much more ambitious launch narrative.
··········
The official benchmark story is strongest in coding, computer use, and knowledge-work evaluations.
OpenAI is publishing visible gains over GPT-5.4, especially in the agentic and workflow-heavy areas that matter most to professional users.
On the coding and agentic side, OpenAI says GPT-5.5 reaches 58.6% on SWE-Bench Pro (Public) compared with 57.7% for GPT-5.4, 82.7% on Terminal-Bench 2.0 compared with 75.1% for GPT-5.4, and 73.1% on Expert-SWE (Internal) compared with 68.5% for GPT-5.4.
On broader workflow and knowledge-work tasks, OpenAI reports 84.9% on GDPval, 78.7% on OSWorld-Verified, and 98.0% on Tau2-bench Telecom, using those numbers to reinforce the idea that GPT-5.5 is stronger for knowledge work, computer-use tasks, and complex multi-step execution.
These are official OpenAI-published benchmarks, which means they are strong and highly relevant to the launch while still remaining vendor benchmark posture rather than a neutral final verdict across the entire market.
........
· The biggest official improvements are in agentic coding and workflow-heavy evaluations.
· GPT-5.5 is also being positioned strongly for knowledge work and computer-use tasks.
· These figures are official OpenAI benchmarks and should be read as vendor-published benchmark posture.
........
Official benchmark highlights for GPT-5.5
Benchmark | GPT-5.5 | GPT-5.4 |
SWE-Bench Pro (Public) | 58.6% | 57.7% |
Terminal-Bench 2.0 | 82.7% | 75.1% |
Expert-SWE (Internal) | 73.1% | 68.5% |
GDPval | 84.9% | — |
OSWorld-Verified | 78.7% | — |
Tau2-bench Telecom | 98.0% | — |
··········
OpenAI is also making a strong speed-and-efficiency argument around GPT-5.5.
The company says GPT-5.5 reaches a higher intelligence level without becoming slower than GPT-5.4 in real-world serving.
OpenAI says GPT-5.5 matches GPT-5.4 per-token latency in real-world serving, which is a very important part of the launch story because major model upgrades often introduce new friction through slower responses or heavier serving costs.
The launch page also says GPT-5.5 uses significantly fewer tokens to complete the same Codex tasks, which lets OpenAI frame the release not only as an intelligence gain but also as an efficiency gain.
That combination matters for ChatGPT because a model that becomes stronger without becoming slower changes the everyday product experience much more directly than a model that improves only on benchmark charts.
··········
The biggest practical change inside ChatGPT is that harder work now requires less hand-holding.
OpenAI is presenting GPT-5.5 as more capable of carrying work forward on its own, especially when tasks are ambiguous, messy, or multi-step.
This is the part of the release that matters most in actual use, because OpenAI is describing GPT-5.5 as a model that can manage ambiguity better, use tools more effectively, and continue pushing through hard tasks with less need for step-by-step steering from the user.
The official positioning also stresses smarter and more concise answers for coding, research, information synthesis, analysis, and document-heavy work, which implies a model that is meant to save not just time on generation, but also time on supervision and correction.
OpenAI says early testers found GPT-5.5 Pro significantly more comprehensive, accurate, relevant, and useful than GPT-5.4 Pro, particularly in business, legal, education, and data science, which reinforces the idea that the company sees this launch as a real change in difficult professional work rather than a shallow incremental refresh.
··········
GPT-5.5 is also replacing older GPT-5-era layers inside ChatGPT’s model structure.
The launch is part of a broader cleanup and replacement cycle, not a simple parallel addition beside older GPT-5 layers.
OpenAI’s Help Center says that as of February 13, 2026, ChatGPT retired GPT-5 Instant and GPT-5 Thinking, and that older chats now run on GPT-5.3 and GPT-5.5 equivalents, which means GPT-5.5 is already part of a wider restructuring of the ChatGPT model stack.
The same support page says GPT-5.5 Thinking will remain available in Legacy Models for 90 days after launch for Plus and Pro users, which makes the transition explicit rather than silent.
That product cleanup matters because it shows GPT-5.5 entering ChatGPT not as a side experiment, but as a central replacement layer in OpenAI’s evolving model structure.
........
· GPT-5.5 is part of a larger replacement cycle inside ChatGPT.
· Older GPT-5-era layers are already being retired or redirected.
· Legacy access remains for a limited transition period.
........
How GPT-5.5 fits into ChatGPT’s evolving model structure
Area | Current official position |
Older GPT-5 Instant / Thinking | Retired earlier in 2026 |
Older chats | Run on GPT-5.3 / GPT-5.5 equivalents |
GPT-5.5 Thinking in Legacy Models | 90 days after launch for Plus / Pro |
··········
Safety is part of the launch story, not a side note.
OpenAI is releasing GPT-5.5 with heightened safeguards and separate red-teaming attention, which shows that the company sees the model as materially more capable.
OpenAI says GPT-5.5 launches with its strongest safeguards to date, which already signals that the company is treating the model as a consequential increase in capability rather than as a routine maintenance update.
The separate GPT-5.5 Bio Bug Bounty strengthens that reading, because a dedicated bug bounty around biological safeguards indicates focused external testing for a model that OpenAI considers advanced enough to deserve targeted scrutiny in high-risk domains.
That safety posture is part of the launch itself.
It is not background noise.
It is one of the ways OpenAI is signaling how seriously it takes the step from GPT-5.4 to GPT-5.5.
··········
GPT-5.5 is a real step beyond GPT-5.4 because the gains reach model quality, product structure, and practical ChatGPT use at the same time.
The release matters because OpenAI is improving the reasoning layer, expanding the premium split, and changing how difficult work feels inside ChatGPT.
GPT-5.5 is already live in ChatGPT and Codex, GPT-5.5 Thinking is now the main advanced reasoning layer in the product, GPT-5.5 Pro gives higher-end users a separate premium tier, and the official benchmark and workflow claims position the model as a real step beyond GPT-5.4 in coding, research, computer use, and difficult multi-step work.
The launch also changes the product in structural ways through plan gating, usage ceilings, legacy-model handling, and a more visible separation between broader advanced access and higher-accuracy premium access.
The API still remains just ahead rather than fully live today, which means the first major impact of GPT-5.5 belongs to ChatGPT itself.
That is what makes the release so important right now.
The model improved.
The product changed with it.
And harder work inside ChatGPT is now being presented as faster, stronger, and more self-sustaining than before.
·····
FOLLOW US FOR MORE.
·····
·····
DATA STUDIOS
·····
