OpenAI's GPT-5.5 is here, and it's no potato: narrowly beats Anthropic's Claude Mythos Preview on Terminal-Bench 2.0

OpenAI launched GPT-5.5 and a higher-end GPT-5.5 Pro, positioning them as stronger than GPT-5.4 for coding, computer tasks, and scientific research.

The model narrowly outperformed Anthropic’s Claude Mythos Preview on Terminal-Bench 2.0, giving OpenAI a new flagship benchmark win.

For now, access is limited to paid ChatGPT and business tiers, with API availability still not announced.

The release could boost advanced coding and workflow automation, but broader developer use will have to wait.