OpenAI has announced a significant model upgrade to its ChatGPT platform: GPT-5.5. The company states that this new version is designed to improve results primarily through enhancements in contextual understanding. OpenAI released a lengthy press release detailing the functionalities and changes incorporated into GPT-5.5.
According to the report, the upgrade is expected to deliver significant improvements across various familiar professional tasks. Specific areas of expected improvement include coding, general computer use, and scientific research.
OpenAI’s data suggests that GPT-5.5 performs overall better when compared to previous versions. An internal AI intelligence index indicates that the model is more efficient. Specifically, the data shows that GPT-5.5 handles tasks of the same difficulty faster than GPT-5.4, utilizing significantly fewer tokens.
This trend of increased efficiency is evident across multiple tasks. In the area of agentic coding, the model’s power has increased due to better reasoning capabilities. Early testing has also pointed to improvements in behaviors critical to real-world engineering work. These improvements include maintaining context across large systems, reasoning through ambiguous failures, checking assumptions using external tools, and managing changes within a surrounding codebase.
Furthermore, OpenAI noted similar gains in computer use, attributing this to the model’s improved ability to understand intent. The overall result cited is a faster model performance when tackling complex assignments. The release also highlighted an improved latency for time-intensive tasks, such as document creation. The new model reportedly completes these actions with better comprehension without sacrificing time.
OpenAI claims that GPT-5.5 can now parse task goals from unstructured or “messy business” contexts and transform them into actionable plans. The new GPT-5.5 upgrade will be available starting today for Plus, Pro, Business, and Enterprise user tiers in ChatGPT and Codex. GPT-5.5 Pro is specifically noted as being available for all these tiers except for Plus.
However, API access to the model is not yet available, reportedly due to the necessary requirement for “different safeguards.” The company reiterated that the remaining aspects of the model are deemed safe, with internal and external teams continuing to test for potential flaws and safety issues.