The operating cost of OpenAI's o3 model may far exceed the initial estimate. The o3 inference model released by OpenAI in December last year has recently been re - evaluated for its operating cost. The latest estimate by the Arc Prize Foundation, the maintainer of the ARC - AGI benchmark test, shows that the model's high - performance...

The Operating Cost of OpenAI's o3 Model May Far Exceed the Initial Estimation

The operating cost of the o3 inference model released by OpenAI in December last year has recently been re-evaluated. The latest estimate by the Arc Prize Foundation, the maintainer of the ARC-AGI benchmark test, shows that the cost of processing a single task for the high-performance version of this model could be as high as $30,000, a significant increase from the initial estimate of $3,000. Notably, the high-performance version of o3 needs to attempt 1,024 times to complete a single ARC-AGI task.

This revision highlights the high cost of current cutting-edge AI models for certain tasks. Although OpenAI has not announced the pricing of o3, the foundation believes that the pricing of its o1-pro model, which is currently the most expensive model of OpenAI, is of reference value. Tests show that the computational volume of the high-performance version of o3 is 172 times that of the low-end version.

TechCrunch

📮Contribute ☘️Channel 🌸Chat

via Tech Circle🎗 Zaihua Channel📮 - Telegram Channel