Some users point out that GPT-5's pricing is lower compared to models of the same level, and also lower than OpenAI's previous version.
OpenAI launched its latest flagship large model GPT-5 last week, with the company's CEO Altman claiming GPT-5 is the world's best model. Just a few days earlier, OpenAI had just released two free open-source models.
OpenAI's strategy of quickly launching multiple products has surprised the industry, and what catches the attention of AI practitioners is that OpenAI has set an extremely competitive price for GPT-5, which may break the pricing system of other companies.
The input cost for top-tier GPT-5 API is $1.25 per million tokens, and the output cost is $10 per million tokens.
This fee is the same as Google's Gemini 2.5 basic subscription price, but Google will charge higher fees after input or output exceeds 200,000 tokens.
Another competitor, Anthropic, has set a higher price for Claude Opus 4.1, with an input token cost of $15 per million tokens and an output token cost of $75 per million tokens.
Google's Gemini 2.5 and Anthropic's Claude Opus 4.1 are close to GPT-5 in performance, but are more expensive, which makes developers highly appreciate GPT-5.
Unable to Cover Costs
Although some users complained after GPT-5's launch that the new model is not as human-friendly as GPT-4o, leading OpenAI to reopen previous models, for developers, GPT-5 is obviously more cost-effective.
Matt Shumer, co-founder and CEO of OthersideAI, said that GPT-5 is cheaper than GPT-4o, which is really great, with the intelligent value created per dollar continuously improving.
Some users directly tout GPT-5 as a "pricing killer" and potentially triggering a price war among other AI companies. However, some tech industry professionals point out that OpenAI's current pricing cannot cover costs, which adds many uncertainties to the future.
The tech industry has invested hundreds of billions of dollars in building data centers and other infrastructure in the US to meet AI demands, but it is still years away from AI creating sufficient profits.
OpenAI has signed a $30 billion annual data center contract with Oracle, but its annual recurring revenue has only recently reached $10 billion. Other large tech companies are taking bigger steps: Meta plans to invest up to $72 billion in AI infrastructure this year; Alphabet, Google's parent company, has reserved $85 billion for capital expenditure this year.
These expenses typically mean that AI model costs will rise rapidly in the coming years. Some industry insiders comment that startups with API needs should not celebrate OpenAI's current low-price strategy too much, as it is likely a temporary measure to seize the market and may involve significant price increases in the future.
This article is from the WeChat public account "Science and Technology Board Daily", author: Ma Lan, authorized by 36kr for publication.
