OpenAI rolls out GPT-5.5 with improved contextual understanding, Plus and up

OpenAI just announced that ChatGPT is getting a model upgrade to GPT-5.5. The company says the model will bring better results because of changes to how it understands context.

OpenAI released another lengthy press release detailing GPT-5.5. The update comes with a few changes over the previous model. It should perform significantly better across various familiar tasks, such as coding, computer use, and scientific research. The company says these improvements are the product of better context understanding.

In comparison across various tasks through data provided by OpenAI, GTP-5.5 appears to carry improvements on the whole. The AI intelligence index paints a picture of a more efficient model. It handles tasks at the same difficulty faster than GPT-5.4, using significantly fewer tokens.

This looks to be the common theme across tasks from a look at the company’s data, with agentic coding seeing an increase in power because of that better reasoning.

Early testing suggests GPT‑5.5 is better at the behaviors real engineering work depends on, like holding context across large systems, reasoning through ambiguous failures, checking assumptions with tools, and carrying changes through the surrounding codebase.

OpenAI says that GPT-5.5 is seeing similar gains in computer use, since it can understand intent better. The result is a faster model when put up against complex tasks. The release also stresses an improvement in latency for document creation and other tasks that generally take a chunk of time. The new model completes these with a better understanding without losing time.

The company says this model should be able to parse out task goals from “messy business” and turn it into an actual plan.

The GPT-5.5 upgrade will be available for Plus, Pro, Business, and Enterprise users in ChatGPT and Codex starting today. GPT-5.5 Pro will be available for all of those except for Plus. APIs don’t appear to be ready because of the requirement for “different safeguards.” The rest of the model is stressed as safe, with internal and external teams testing for flaws and safety issues.