Competition to release stronger AI models is accelerating, and just weeks after the release of GPT-5.4, OpenAI has introduced GPT-5.5, pointing to expanded safeguards in the new model.
GPT-5.5 is being rolled out to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex, while GPT-5.5 Pro is available to Pro, Business, and Enterprise users in ChatGPT. The models are expected to be made available through the API soon. The company said it is working with partners and customers to address safety and security requirements for deployment at scale.
Stronger Safeguards and Refined Controls
“We are releasing GPT‑5.5 with our strongest set of safeguards to date, designed to reduce misuse while preserving access for beneficial work,” the company said in the announcement. “We evaluated this model across our full suite of safety and preparedness frameworks, worked with internal and external redteamers, added targeted testing for advanced cybersecurity and biology capabilities, and collected feedback on real use cases from nearly 200 trusted early-access partners before release.”
OpenAI says cyber-specific safeguards introduced with GPT-5.2 have been refined in subsequent releases. For GPT-5.5, the company implemented tighter controls around higher-risk activity, including restrictions on sensitive cybersecurity-related requests and protections against repeated misuse. These measures are supported by monitoring systems and authenticated access controls, and were tested with external experts to assess their robustness. The updated safeguards also apply stricter limits to cyber workflows considered more likely to be abused, while maintaining access for legitimate development and security use cases.
Tests Reveal Strengths and Limitations
Senior engineers who tested the model found GPT-5.5 to be stronger than GPT-5.4 and Claude Opus 4.7 in reasoning and autonomy, including identifying issues in advance and anticipating testing and review needs without explicit prompting. “GPT-5.5 is noticeably smarter and more persistent than GPT-5.4, with stronger coding performance and more reliable tool use,” noted Michael Truell, CEO at Cursor.
Irregular, a security lab focused on testing advanced AI systems, evaluated GPT-5.5 and found meaningful performance improvements across a range of task categories. “The improvement is most relevant for novice and moderately skilled operators, while also offering targeted assistance to highly skilled experts on precise, narrow subtasks,” researchers wrote. “This proficiency is particularly effective in streamlining workflows, especially for vulnerability research and exploitation when the scope of the task is well defined.”
Researchers note that, in some cases, the model was able to perform complex cyber tasks requiring niche knowledge that most expert cyber operators would not possess. These results suggest the model has the potential to alleviate existing bottlenecks in scaling cyber operations through the automation of discovering and exploiting operationally relevant vulnerabilities. However, researchers emphasize that, despite these advancements, limitations persist. “We still see constraints in translating these capabilities to real-world scenarios due to limitations in areas such as operational security.” “Consistent with previous assessments, these outcomes should be interpreted as a measure of its capabilities for assisted reasoning, not as a reflection of its efficacy in real-world attack scenarios,” they concluded.
Performance Gains and Pricing
According to OpenAI, GPT-5.5 shows stronger performance in tasks such as coding, computer use, knowledge work, and early-stage scientific research, which require reasoning across context and multi-step execution. It maintains similar per-token response speeds to GPT-5.4 while delivering higher overall performance. The model also uses fewer tokens to complete comparable coding tasks, which could improve efficiency in practical use.
GPT-5.5 is priced higher than GPT-5.4, but the company claims “it is both more intelligent and much more token-efficient.” OpenAI stated: “GPT-5.5 is an incremental but important step towards AI that can solve some of the world’s toughest challenges like cybersecurity.”
With expanded safeguards and stronger capabilities, GPT-5.5 represents a significant step forward in balancing power and safety, though its real-world attack potential remains limited. The model is now available to select user tiers, with broader API access expected shortly.
Source: Help Net Security News