GPT-5.5 

Share

Latest news on GPT-5.5, OpenAI's most advanced AI model, covering ChatGPT updates, agentic coding, AI safety, benchmarks, and the race toward AGI.

GPT-5.5 is OpenAI's latest large language model (LLM), released on 23 April 2026. Described by the company as its "smartest and most intuitive" model yet, it is designed for complex, multi-step tasks including agentic coding, computer use, knowledge work, and early scientific research. OpenAI co-founder and president Greg Brockman called it a significant step toward a new kind of computing and toward the company's vision of an AI "super app".

The model launched for Plus, Pro, Business, and Enterprise users in ChatGPT and Codex, with API access following a day later. GPT-5.5 Pro, a higher-compute variant, rolled out to Pro, Business, and Enterprise subscribers. A lighter version, GPT-5.5 Instant, became available to free-tier users on 5 May 2026. Despite being more capable than its predecessor GPT-5.4, the model matches it on per-token latency and uses fewer tokens to complete equivalent tasks in Codex, making it both faster and more cost-efficient in practice.

Safety has been a central point of debate around GPT-5.5. The UK's AI Security Institute rated it the world's most capable model on individual cyber tasks, finding it able to complete a complex, multi-step corporate network attack simulation. Critics argue that OpenAI alone should not decide whether such a model is safe to deploy publicly, calling for independent regulatory oversight. OpenAI maintains it has applied its strongest safeguards to date, including red-teaming for advanced cybersecurity and biology capabilities, and a vetted "Trusted Access" programme for sensitive use cases.

GPT-5.5 arrives amid a broader debate about the pace and governance of frontier AI development. OpenAI's release cadence has accelerated sharply, from GPT-5 in August 2025 to GPT-5.2 in December 2025, GPT-5.4 in March 2026, and GPT-5.5 the following month. Each release has intensified competition with Anthropic and Google DeepMind, whose Gemini 3.1 Pro and Claude models serve as recurring benchmarks for comparison. OpenAI's own data shows GPT-5.5 outperforming both on several key evaluations, though independent reviewers have produced more mixed results.

Beyond benchmarks, GPT-5.5 is generating interest for its potential real-world impact. OpenAI has pointed to applications in drug discovery and scientific research, areas where reasoning across large bodies of information over extended sessions could accelerate progress. NVIDIA reported that more than 10,000 of its staff used early access to the model through Codex, spanning engineering, legal, finance, and operations roles. This signals a shift in how frontier AI is being positioned: not just as a developer tool, but as infrastructure for general knowledge work across organisations.

The NewsNow feed on GPT-5.5 is your one-stop source for the most relevant headlines as they break, tracking OpenAI announcements, ChatGPT updates, AI safety developments, and the wider LLM landscape. Whether you follow AI research, work in tech, or are simply curious about where artificial intelligence is heading, this feed keeps you informed with the latest coverage.