OpenAI GPT 5.5 and 5.5 Pro launch with $25,000 bounty

OpenAI has unveiled its latest AI models, GPT 5.5 and GPT 5.5 Pro, powering its ChatGPT chatbot service and AI APIs. Both GPT models are improved in intelligence when compared to GPT 5.4 as well as its competitors, Claude Opus 4.7 and Gemini 3.1 Pro, but they also come with increased risks. GPT 5.5 and 5.5 Pro are available to paying subscribers of ChatGPT, and they will be available soon to API users.

The latest AIs from OpenAI continue to improve in their ability to solve challenging academic problems and use computers as tools. The improvements also bring an increased know-how of the protocols needed to create biological threats and the methods needed to successfully hack networks and systems. Often, the performance of GPT 5.5 and GPT 5.5 Pro in these areas is better than that of competing AI models Opus 4.7 and Gemini 3.1 Pro. Lately, users of Claude models have found more insecure code being generated by the Anthropic AIs.

Although added safeguards have been implemented due to the high risk rating of OpenAI's latest models, the company has opened a Bio Bug Bounty for GPT‑5.5 that will award $25,000 to those who can jailbreak GPT 5.5 in Codex Desktop when faced with a five-question biosafety challenge. Interested readers can apply between April 23 and June 22, 2026.

Anthropic Claude Mythos is already so good at finding cybersecurity bugs that the company will not release the AI due to its enormous national security risk. Its less-capable, publicly available Claude Code has already been used to crack FreeBSD.

Readers interested in using GPT locally can download the older, open-source GPT-OSS model from Hugging Face and run it on a PC with an Nvidia GPU that has 16 GB of memory or greater (like this 5090 GPU on Amazon).