AI

An optimised inference stack means lower costs.
OpenAI is rolling out o3-pro, a “more reliable” version of its top-end o3 model, with prices slashed for developers as it seeks to keep up with a maturing market and accessibility demands.
The company revealed prices for the o3-pro model would be 87% lower then those for the o1-pro model it is replacing, while prices for the lower o3 version would also be dropped by 83%.
It said savings had been made by optimising the inference stack that serves o3 and claimed the price drops made it “one of the most cost-efficient frontier reasoning models available.”
The release of o3-pro comes just two months after the standard o3 model was launched in April and offers more reliable and well-reasoned responses to user inputs, at the cost of slower response times.
OpenAI claimed the new model was consistently rated higher than o3 during “expert evaluations” and showed gains on coding, maths and reliability benchmarks over its predecessors.
Developer prices for o3-pro, now available for OpenAI’s Pro and Team users, will cost $20 per 1M tokens input and $80 per 1M tokens output, while o3’s new prices are now $2 per 1M tokens input and $8 per 1M output, making it cheaper than OpenAI's GPT-40.
While the new prices are significantly lower then the $150 input and $600 output costs seen for o1-pro, they still put o3-pro at the higher end of the market, with DeepSeek R2 at just $0.14 per 1M tokens input and $0.55 per 1M output, and Google’s Gemini 2.5 Pro, currently in preview, with top prices at $2.50 per 1M tokens input and $15 per 1M output.
OpenAI still dominates the LLM and Generative AI market but its price cuts come after fluctuations in its market share, dropping from 50% to 34% of the Enterprise LLM market in 2024, according to Menlo Ventures, after increasing competition from up and comers such as Anthropic, and big tech players including Google and Meta.
However, its B2C offering still remains very strong, with website traffic site Similarweb claiming OpenAI is responsible for almost 80% of the traffic to Generative AI tools online, a number that has risen over the last six months.
Interviews, insight, intelligence, and exclusive events for digital leaders.
No spam. Unsubscribe anytime.