OpenAI has unveiled its latest models, GPT-5.5 and GPT-5.5 Pro, enhancing capabilities in coding and complex task execution. This release comes shortly after Anthropic launched Opus 4.7, marking a competitive moment between the two AI firms, which are rapidly evolving their offerings in tandem.
Simon Willison, a developer who gained early access to GPT-5.5, noted its speed and effectiveness but encountered a significant drawback: the absence of API access. This limitation hindered his ability to conduct standard tests, such as generating an SVG of a pelican on a bicycle, a challenging prompt designed for evaluating model performance in unfamiliar tasks.
Despite initial setbacks, Willison managed to work around the API issue by creating a plugin using a Codex backdoor. While he found the default performance of GPT-5.5 lacking compared to GPT-5.4, adjustments for increased reasoning time yielded improved results, albeit with longer response times and higher token consumption. Other users have noted a reduction in the amount of guidance required for effective outputs from the new model.