OpenAI priced GPT-5 so low, it may spark a price war

For the second time this week, OpenAI was surprised by its latest flagship model, GPT-5, to be launched, only a few days after releasing two new freely available models under an open source license.

OpenAi CEO Sam Altman went so far that he called GPT-5 ‘the best model in the world’. That can be proud or hyperbole, because Maxwell Zeff from WAN reports that GPT-5 is only performing slightly better than other leading AI models from Anthropic, Google DeepMind and Xai on some important benchmarks and somewhat delays on others.

Yet it is a model that performs well for a wide range of applications, in particular coding. And, as Altman noted, an area where it undoubtedly competes well, is the price. “Very happy with the prices we can deliver!” He tweeted.

The GPT-5 API at the highest level costs $ 1.25 per 1 million tokens of input, and $ 10 per 1 million tokens for output (plus $ 0.125 per 1 million tokens for input in the cache). This price reflects Google’s Gemini 2.5 Pro Basic Subscription, which is also popular for coding -related tasks. However, Google counts more if inputs/outputs cross a heavy threshold of 200,000 instructions, which means that the most consumption-heavy customers ultimately pay more.

But OpenAi Opus 4.1 of Anthropic Echt undermines $ 15 per 1 million input tokens and $ 75 per 1 million output sticks. (Anthropic, however, offers major discounts for fast caching and batch processing – storing/reusing prompts and the processing of multiple requests together.)

The Anthropic model is extremely popular with programmers, as well as a choice within the popular coding assistant cursor and for feeding his own such assistant, Claude Code. (Notice that Cursor offered GPT-5 as an option minutes after it was announced.)

Developers who have early access to GPT-5 can be appointed the prizes. Simon Willison, one of the developers on display in the launch video of OpenAiwrites in his judgement: “The prices are aggressive competitive With other providers. “

But GPT-5 is also competitively priced with GPT-4O. Other Ideai’s co-founder and CEO, Matt Shumer (maker of Hyperwrite), writes That GPT-5 “is cheaper than GPT-4O, which is fantastic. Intelligence per dollar continues to increase.”

Some on X mentioned the costs of OpenAI for the model “A price killer“While others are on Hacker News Offer a similar praise.

Will competitors such as anthropic follow? Will Google – who undermine the prices – will become even more affordable? If so, we could witness the start of a long-awaited LLM prize war.

There is no doubt that a price war would be welcome. For example, the underlying economy of suppliers of atmospheric tools is pretty shaky because of the high and unpredictable costs they have to pay for models, as Marina Temkin of WAN reports. And there are also countless startups that build on top of AI models.

Silicon Valley has hoped that the LLM-PRIJS-Performance ratio will ultimately improve, together with the conclusion costs. But it seemed that such an equalization could be gone for years, because the technical industry is investing hundreds of billions to build data centers and infrastructure to support the growing AI question.

OpenAi itself has a $ 30 billion contract with Oracle for capacity, if it only recently achieved an annual return of $ 10 billion. In the meantime, Meta is planning to spend up to $ 72 billion on AI infrastructure in 2025, and has an alphabet set aside $ 85 billion for capital expenditure in 2025, powered by AI needs. In the light of such enormous costs, the costs usually go in a way: up.

In view of such investments, it can be too early for startups looking at their rising model API accounts to rejoice from the only step from OpenAI to lower prices.

But this week, OpenAi threw the glove down to exert the prices not only once but twice pressure. We will see if others follow.