OpenAI releases GPT-5.5, reclaiming top spot in publicly available AI — beats Anthropic's Claude on coding benchmark

OpenAI unveiled GPT-5.5 today after months of development (internal codename: 'Spud'). The model is now available through ChatGPT and OpenAI's API, positioned as the company's strongest offering to date and narrowly outperforming Anthropic's private Claude Mythos Preview model on Terminal-Bench 2.0 (a standardized test that measures AI reasoning and coding ability).

GPT-5.5 shows measurable gains in coding tasks — the most concrete differentiator OpenAI highlighted. This means software engineers using ChatGPT or the API will see fewer errors when asking the AI to write or debug code, and faster turnaround on multi-step programming problems, especially on projects where accuracy matters (data pipeline validation, security-critical functions).

For business users and developers evaluating which AI assistant to pay for, the choice just shifted: if coding quality and speed are your primary need, GPT-5.5 now edges out Claude and Google's alternatives. For companies building applications on APIs, this may justify staying with or switching to OpenAI's platform — though expect Anthropic and Google to respond with their own upgrades within months.