error-rate-reduction

OpenAI partners with Databricks to release the GPT-5.5 model, achieving a 46% reduction in error rate in agent frameworks, becoming the only model to exceed 50% on benchmarks, with significant improvements in parsing quality and function calling capabilities.