We first saw the makers of ChatGPT share the latest o3 Reasoning AI Model in December. Since then, the organization has chosen to partner with the makers of ARC-AGI to display o3’s offerings.
Now, months later, it’s safe to say the results were revised and they’re slightly less impressive than what they hoped for. In the past, the administrators of ARC-AGI shared the approximate computing costs for the o3 model. The firm had originally estimated that the best-performing configuration cost nearly $3000. This is to only solve one ARC-AGI issue.
Now, the cost is actually calculated to be nearly $30k for each task. The latest revision is notable as it illustrates how expensive today’s most sophisticated AI models are for simple tasks. While OpenAI isn’t pricing the product just yet or releasing it, experts feel pricing is quite a reasonable proxy.
The o1-pro is definitely the company’s most expensive model right now. It’s quite a closer comparison to the actual price for o3, thanks to the great amount of test time for computing utilized. Still, it’s a proxy for now, and they’ve got o3 labeled as previews on leaderboards to display the uncertainty linked to actual pricing today.
There is a high cost for the o3 high that wouldn’t be out of this question. Remember, it’s using a high quantity of computing resources. We’re talking 172 times more computing than the lower o3, which is the lowest configuration for tackling ARC-AGI.
Other than that, there are plenty of speculations surrounding the price plans for OpenAI rolling out enterprise clients. Towards the start of March, we saw The Information share how the company plans to charge nearly $20k every month for special AI agents such as software developers.
People might not agree with the strategy used by OpenAI’s priciest models, costing below what common contractors would command. But as a top AI researcher shared through a post on X, these models aren’t too efficient. For instance, the o3 needs up to 1024 attempts for every task across the ARC-AGI range to get the best score.
Still, it’s all going to depend on pricing, but seeing it fall short of people’s expectations regarding its capabilities and costs might impact its popularity.
Image: DIW-Aigen
Read next: Wikipedia Pays Price of AI Boom: Company Struggling From Rising Costs Due to Bots Scraping Its Articles
Now, months later, it’s safe to say the results were revised and they’re slightly less impressive than what they hoped for. In the past, the administrators of ARC-AGI shared the approximate computing costs for the o3 model. The firm had originally estimated that the best-performing configuration cost nearly $3000. This is to only solve one ARC-AGI issue.
Now, the cost is actually calculated to be nearly $30k for each task. The latest revision is notable as it illustrates how expensive today’s most sophisticated AI models are for simple tasks. While OpenAI isn’t pricing the product just yet or releasing it, experts feel pricing is quite a reasonable proxy.
The o1-pro is definitely the company’s most expensive model right now. It’s quite a closer comparison to the actual price for o3, thanks to the great amount of test time for computing utilized. Still, it’s a proxy for now, and they’ve got o3 labeled as previews on leaderboards to display the uncertainty linked to actual pricing today.
There is a high cost for the o3 high that wouldn’t be out of this question. Remember, it’s using a high quantity of computing resources. We’re talking 172 times more computing than the lower o3, which is the lowest configuration for tackling ARC-AGI.
Other than that, there are plenty of speculations surrounding the price plans for OpenAI rolling out enterprise clients. Towards the start of March, we saw The Information share how the company plans to charge nearly $20k every month for special AI agents such as software developers.
People might not agree with the strategy used by OpenAI’s priciest models, costing below what common contractors would command. But as a top AI researcher shared through a post on X, these models aren’t too efficient. For instance, the o3 needs up to 1024 attempts for every task across the ARC-AGI range to get the best score.
Still, it’s all going to depend on pricing, but seeing it fall short of people’s expectations regarding its capabilities and costs might impact its popularity.
Read next: Wikipedia Pays Price of AI Boom: Company Struggling From Rising Costs Due to Bots Scraping Its Articles
