The company has unveiled its flagship AI model. GPT-5.5 understands user intent, plans work autonomously, and completes tasks from start to finish.
OpenAI has unveiled its flagship AI model, GPT-5.5. The new model is positioned as a “new level of intelligence for real work and agent management.” The model understands user intent, plans work autonomously, and completes tasks from start to finish.
Hot topic: Bitcoin on April 24—BTC Price Stuck in $77,500–$78,500 Range
GPT-5.5 is already available to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex.
Contents
Programming, Science, and Office Work: Where the New Model Excels
GPT-5.5 is particularly effective at agentic programming, computer control, knowledge work, and early-stage scientific research. Here are some key test results:
- Terminal-Bench 2.0 (command line): 82.7%
- SWE-Bench Pro (GitHub problem-solving): 58.6%
- GDPval (knowledge work across 44 professions): 84.9%
- OSWorld-Verified (computer environment tasks): 78.7%
- Tau2-bench (customer service): 98.0%
On GeneBench (genetics and quantitative biology), the model showed significant improvement over GPT-5.4. On BixBench (bioinformatics), GPT-5.5 outperformed all competitors with published scores.
Read also: Pope Criticizes AI in Posts Written by… Artificial Intelligence
Why GPT-5.5 Doesn’t Slow Down as Intelligence Grows
OpenAI stated that GPT-5.5 delivers a leap in intelligence without sacrificing speed. The model matches GPT-5.4 in token latency while demonstrating a much higher level of reasoning. It also uses “significantly fewer” tokens when working in Codex.
OpenAI President Greg Brockman called the release a step toward “more agentic and intuitive computing.” The new version also brings the company closer to launching a “super app” that will combine ChatGPT, Codex, and an AI Browser.
Read also: ‘Too Powerful’—Anthropic Restricts Access to Mythos AI Model
The Cost: GPT-5.5 Is More Expensive but Cheaper to Use
The base version of GPT-5.5 via API will cost $5 per 1 million input tokens and $30 per 1 million output tokens, compared to $2.5 and $15 for GPT-5.4. The Pro version costs $30 and $180. The context window is 1 million tokens.
However, OpenAI CEO Sam Altman stated that cost savings from using fewer tokens will offset the price increase. GPT-5.5 completes the same tasks in Codex at a lower cost.
For comparison: Xiaomi MiMo v2.5 Pro costs $1 and $3 per million tokens. Minimax M2.7 costs $0.30 and $1.20.
Read also: ARK Invest sees AI infrastructure market nearing $1.5 trillion by 2030
