OpenAI has released its new AI models called GPT-5.5 series. It says these are smarter and made for real-world tasks like coding and deep knowledge work. The update was announced on Thursday, and paid users of ChatGPT and Codex got early access.
The company claims these models perform better than competitors like Anthropic’s Claude Opus 4.7 and Google’s Gemini 3.1 Pro in important tests. This shows a big improvement in how AI can handle difficult tasks on its own.
The GPT-5.5 series has three main types for different needs. There is the standard GPT-5.5, the GPT-5.5 Thinking version for step-by-step reasoning, and the advanced GPT-5.5 Pro for heavy tasks. Right now, GPT-5.5 and its Thinking version are available to ChatGPT users on Plus, Pro, Business, and Enterprise plans.
However, the Pro model is only available for Business and Enterprise users, not regular Pro users yet. In Codex, the base GPT-5.5 is available for Go, Plus, Pro, Business, Enterprise, and Edu users. It includes a large 400K token context window, which allows it to handle a lot of information at once, and also has a Fast mode for quicker performance.
Also Read: Oppo Find X9 Ultra Teardown Reveals Powerful Design, Advanced Cameras, and Massive Battery
One key feature of these models is agentic coding. This means the AI works like a smart teammate. It can plan, write code, debug, and fix problems without needing constant instructions. In the Terminal-Bench 2.0 test, GPT-5.5 scored 82.7%, beating other models. It also scored 73.1% on SWE-Bench Pro for long coding tasks.
In real use, Codex with GPT-5.5 can improve code, test changes in large projects, find unclear errors, verify results with tools, and continue working even in complex situations. OpenAI says this also reduces the number of tokens used, making it more efficient.
Apart from coding, GPT-5.5 is strong in knowledge-based tasks. It understands what users mean even if instructions are not clear. It can find information, test ideas, analyze results, and decide the next steps. This is useful in areas like science, technology research, and office work such as creating spreadsheets or reports.
Greg Brockman said the model can handle unclear problems on its own and decide what to do next with less guidance. This is important for building future AI agents that can use computers independently. Users can give it complex tasks involving multiple steps, and it can plan, check its work, and continue without much supervision.
For developers and businesses using APIs, GPT-5.5 will be available soon through Responses and Chat Completions. The pricing is $5 per million input tokens and $30 per million output tokens (around Rs. 471 and Rs. 2,828 in India). The Pro version is more expensive, costing $30 for input and $180 for output per million tokens.
Also Read: iPhone 18 Pro Max Leak Reveals Bigger Cameras and Refined Design Ahead of 2026 Launch
There are also options like Batch and Flex at half price, and Priority at 2.5 times the standard cost. Codex Fast mode runs 1.5 times faster but costs 2.5 times more. Even though prices are higher than GPT-5.4, the model uses fewer tokens, which helps keep overall costs lower.
This release comes soon after GPT-5.4, showing how quickly the AI industry is moving, with companies like Anthropic releasing updates frequently. OpenAI built GPT-5.5 with strong safety features and trained it using Nvidia GPUs. The focus is on making it efficient for coding, research, and AI agents.
Early feedback shows it can work well with office applications on computers, generate images in Codex, and support multiple languages better. This makes it very useful for professionals and creators. As API access expands, more applications and AI agents are expected to use this technology soon.