OpenAI Releases GPT-5.5 With Advanced Agentic Capabilities
OpenAI introduces GPT-5.5, a new frontier model designed for agentic coding, computer use, and scientific research. The model features a massive 1M context window and achieves high scores on abstract reasoning benchmarks.
OpenAI launches GPT-5.5 as its newest frontier model, targeting advanced use cases like agentic coding, computer use, professional knowledge work, and early scientific research. This release improves upon the previous GPT-5.4 model while maintaining the same per-token latency in real-world serving environments. Developers access the model through the Responses and Chat Completions APIs.
The new model boasts an impressive 1 million token context window for API requests and a maximum output capacity of 128K tokens. GPT-5.5 operates more efficiently on Codex tasks by using fewer tokens, and it supports xhigh reasoning during evaluations. The pricing sits at $5.00 per million input tokens and $30.00 per million output tokens.
Performance benchmarks show GPT-5.5 achieving a 95% score on the ARC-AGI-1 abstract reasoning test and an 85% on the newer ARC-AGI-2 visual reasoning puzzles. The model also demonstrates strong capabilities in web comprehension with an 84.4% score on BrowseComp and solid performance in cybersecurity challenges. OpenAI currently restricts access to ChatGPT, Codex, and the official OpenAI API.