#GPT-5.5

Databricks says it is making GPT-5.5 available for customer agent workflows after the model surpassed 50% accuracy on the company’s OfficeQA Pro benchmark and reduced errors by 46% versus GPT-5.4.

Researchers at Carnegie Mellon University say a new benchmark measuring real V8 exploitation found Anthropic's Claude Mythos Preview well ahead of OpenAI's GPT-5.5, though at far higher cost.

A reported OpenAI update would replace ChatGPT’s default model with GPT-5.5 Instant, aiming to make responses feel smarter and more accurate while reducing gratuitous emoji use.

New results from the UK’s AI Security Institute suggest OpenAI’s GPT-5.5 performs at roughly the same level as Anthropic’s heavily restricted Mythos Preview on key cybersecurity evaluations.

OpenAI launched GPT-5.5 on April 23 and positioned it as its most capable agentic AI model yet, emphasizing real-world work and agent use cases while charging roughly double the API price.

OpenAI is offering up to $25,000 for a universal jailbreak that defeats a five-question biology safety challenge in GPT-5.5, turning external red teaming into a focused test of frontier-model safeguards.

OpenAI says GPT-5.5 is built to handle longer, tool-using workflows with less supervision, extending the company’s push toward agent-style AI systems for coding, research, and data work.

OpenAI’s GPT-5.5 has moved to the top of a major benchmark ranking and appears more token-efficient than its predecessor, but reporting cited in the source says the model still hallucinates at a high rate.