GPT-5.5 goes live with agentic features for complex goals

OpenAI rolled out GPT-5.5 on April 23, saying the model is built to handle messy, multi-step work across coding, research, spreadsheets, and software tools. (openai.com) OpenAI said GPT-5.5 is going to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex, while GPT-5.5 Pro is going to Pro, Business, and Enterprise tiers. API access for GPT-5.5 and GPT-5.5 Pro followed on April 24. (openai.com) The company described the model as “agentic,” meaning it is designed to plan a task, use tools, check its own work, and keep going with less step-by-step supervision. OpenAI’s system card says the model asks for less guidance and uses tools more effectively than earlier versions. (openai.com 1) (openai.com 2) On OpenAI’s published benchmarks, GPT-5.5 scored 82.7% on Terminal-Bench 2.0, 78.7% on OSWorld-Verified, 55.6% on Toolathlon, and 84.4% on BrowseComp. OpenAI also said GPT-5.5 matches GPT-5.4 per-token latency in real-world serving. (openai.com) OpenAI tied the release to work done over time, not just one-shot answers. The company said GPT-5.5 uses significantly fewer tokens on the same Codex tasks and gathered feedback from nearly 200 trusted early-access partners before release. (openai.com) Four days later, Amazon Web Services said Amazon Bedrock would offer OpenAI models, Codex, and Bedrock Managed Agents in limited preview. AWS said the service lets customers run OpenAI systems with Identity and Access Management, PrivateLink, encryption, and CloudTrail logging. (aws.amazon.com) AWS said Managed Agents are powered by OpenAI models and the OpenAI agent harness, with each agent getting its own identity and action logs while running inside a customer’s environment. AWS also said usage can count toward existing cloud commitments. (aws.amazon.com 1) (aws.amazon.com 2) OpenAI also published Symphony on April 27 as an open-source orchestration spec for Codex. The company said Symphony turns an issue tracker such as Linear into a control plane that assigns open tasks to dedicated coding agents. (openai.com) (github.com) In that post, OpenAI said Symphony produced a 500% increase in landed pull requests on some teams after engineers stopped juggling three to five interactive Codex sessions and let agents pull work from the task board. The GitHub repository is published under the Apache 2.0 license and is labeled an engineering preview for trusted environments. (openai.com) (github.com) The release sequence put a new model, an orchestration spec, and a managed runtime into production channels within six days. OpenAI’s pitch is that GPT-5.5 should do more of the computer work itself, while AWS is offering the enterprise wrapper around it. (openai.com 1) (openai.com 2) (aws.amazon.com)