OpenAI Launches GPT-5.5: Token Cost Reduced to 1/35, Output Efficiency Increased by 50x

OpenAI recently launched the GPT-5.5 model, its most intelligent AI model to date, excelling in Agentic Coding, computer usage, and scientific research. Compared to previous versions, it can complete ambiguous tasks with less guidance, autonomously plan multi-step operations, and perform remarkably in code debugging and data analysis.

In terms of performance, GPT-5.5 achieves breakthroughs in cost and efficiency: the cost per million tokens is reduced to 1/35 of the previous generation, and the output per megawatt token increases by 50x. Partners like NVIDIA report that debugging time has been cut from days to hours.

Currently, the model is available to Plus and Pro users, with the API version coming soon. Input pricing is $5 per million tokens (Pro version $30), bringing more efficient AI experiences to businesses and individual users.