OpenAI's New Model Costs 1,000$ Per Prompt

Looks like human brainpower comes at an extreme premium

OpenAI’s latest advancement in AI, the O3 model is making new waves with its incredibly complex capabilities—but at an equally incredible cost.

Designed to “reason” more effectively through a technique called test-time compute, O3 takes extra time to evaluate and explore multiple possibilities before delivering an answer. This approach allows it to handle complex prompts with greater accuracy, achieving a groundbreaking 87.5% score on the ARC-AGI benchmark — nearly triple its predecessors.

A steep cost

this leap in performance doesn’t come cheap. In its most powerful mode, O3 reportedly costs over $1,000 per task in computing power. Even its more economical “low-compute” mode costs around $20 per task, while seemingly moderate, it is a drastic increase compared to earlier models like O1, which cost less than $4 per task.

This raises questions about the scalability and affordability of AI advancements. O3 proves that AI gains are far from finished, the industry must grapple with the steep costs that come with this level of performance.

As OpenAI continues to refine its models, cost-performance improvements are anticipated in the coming months and years. For now, O3 remains out of reach for the public, though a “mini” version is slated for release in January.

At Frontier, we’ll be keeping a close eye on how these developments reshape the AI landscape—and what they mean for businesses and users alike. Stay tuned for more updates in the world of cutting-edge tech.