It has been a busy few days (extremely busy) at OpenAI, which appears to have definitively answered the “red code” activated by its CEO Sam Altman at the end of last year. In the same week that we saw the launch of Images 2.0 and Workspace Agents, a new version of GPT has arrived: 5.5.
And the company seems very excited about it.
“What is truly special about this model is everything it can do with less help. It can analyze an unclear problem and determine what needs to happen next. This model represents a real step forward toward the kind of computing we expect in the future. However, it is only one step, and we expect many more,” explained OpenAI co-founder and president Greg Brockman during the presentation.
Specifically, the company seems very proud of this new model’s ability to solve complex problems made up of multiple related subtasks. The main idea behind the launch announcement is that GPT 5.5 can plan, use different tools, verify its work, and keep moving forward until it reaches the core solution to the assigned problem, without needing the subtasks to be explicitly specified.
And it does all of this “often achieving higher-quality results with fewer tokens and fewer retries.”
According to the published metrics, the new model stands out for a major leap in coding and computer control. OpenAI has presented metrics placing it at the forefront, managing to resolve real GitHub issues in 58.6% of cases and reaching 82.7% accuracy in console tasks and tool coordination. In addition, the model can operate software directly in real computing environments, allowing it to manage office workflows such as creating documents, spreadsheets, and presentations, behaving more like an executive assistant than a simple chatbot.
It makes perfect sense that GPT 5.5 was unveiled in the same week as the arrival of Workspace Agents, since this new model clearly seems designed to power the AI agent boom. As we mentioned earlier, the model’s integration with collaborative tools allows AI not only to write, summarize, or analyze information, but also to move forward with longer tasks: searching for data, using tools, checking results, and turning disorganized materials into documents, spreadsheets, presentations, or work plans.
GPT-5.5 is smarter, but also more efficient. OpenAI explains that it has improved token generation speed by 20% and is more efficient, especially in programming tasks, where it can reduce token usage, lowering operating costs compared with competing models. This is complemented by the introduction of GPT-5.5 Pro, a higher-end version for demanding environments in legal, scientific, and data analysis sectors, where precision and complex reasoning are critical.
Finally, OpenAI stresses that this model arrives with its strictest safety systems to date, reinforcing protections in sensitive areas such as cybersecurity and biology. The idea is for AI to be a massive productivity tool, but a safe one as well—something especially important amid the controversy surrounding Anthropic’s Mythos model, a model that has been considered a global security risk
The standard GPT-5.5 model is now available to ChatGPT Plus, Pro, Business, and Enterprise users. The GPT-5.5 Pro version, focused on maximum precision and complex tasks, is reserved for the Pro, Business, and Enterprise tiers. As for its integration with Codex, it will feature a 400,000-token context window for education and enterprise plans, including a fast mode that increases generation speed in exchange for a higher cost.
As for the API, developers will be able to access GPT-5.5 with a one-million-token context window. Base pricing will be $5 (input) and $30 (output) per million tokens, while the Pro version will raise those rates to $30 and $180, respectively. Although the unit price is higher than in previous versions, the model uses fewer tokens to complete the same tasks, which balances the final cost for users thanks to its greater intelligence and optimization.
Image: ChatGPT
Your email address will not be published. Required fields are marked *
Δ