OpenAI o3 has achieved significant success with its performance among artificial intelligence models by pushing the limits. However, the high costs behind this success seem to limit its usage areas. o3, developed with a method called ‘test-time compute’, will redefine the potential of artificial intelligence models by using unprecedented processing power. Here are the details…

The cost of a single mission with OpenAI o3 is $1000!

OpenAI’s o3 model achieved a record success of 88 per cent in the ARC-AGI test. This test measures how close artificial intelligence comes to human-level general skills. O3 attracted attention by far surpassing the 32 per cent score of OpenAI’s previous model o1. However, this success comes with a serious transaction cost. So much so that the cost of a task for o3’s peak performance exceeds $1,000.

The o3 model is currently targeting deep-pocketed users such as large organisations or governments, rather than individual users. While OpenAI offers $200 subscription plans for the o1 model, it is working on plans that can go up to $2,000 for models such as o3. However, these costs are likely to make it difficult for the AI model to be adopted by the masses.

Experts think that the o3 model proves that the scaling laws in artificial intelligence models are still valid. However, if this success is not supported by technological advances such as more efficient AI chips, it will not be possible to undercut the costs. Investors such as Andreessen Horowitz state that startups developing chips in this area will play a more important role in the future.

So, can these developments make AI models accessible to everyone? What do you think about this? Don’t forget to share your opinions in the comments!