He GPT-5 launch It has been, in broad strokes, disappointing. Openai needed this model with this model bigger in the history of AIbut we have encountered a model that improves, but not spectacularly. And yet, it is achieving something that is more important than it seems: to convince companies.
Companies <3 GPT-5. Criticisms have surrounded this launch, but the truth is that the model seems to be capturing more and more companies. Among them are Cursor, Vercel, Jetbrains and Factorythat They argue which have turned GPT-5 into their default model into several products because their implementation is faster and offers better results in complex tasks. Aaron Levi, CEO of Box, said in CNBC that GPT-5 is “a great advance” and has a level of reasoning that other models cannot match. P
From the fact to the fact. The ads of those responsible for these companies in their respective blogs, of course, give very positive visions, but it must be taken into account that they are statements that coincided with the launch and that among other things are part of the business dynamics: the companies support each other because that allows to strengthen alliances for the future. But those same companies that speak of GPT-5 are by agnostic necessity: they are catwalks for a large number of models that users then use to decide which ones are best for them at all times.
Cursor, for example, allows you to use as a base Claude Opus 4.1GPT-5, but also another large number of AI models that can be more interesting for users for both performance and price. The statements of anonymous sources close to these companies are promising for Openai, but not definitive.
Anthropic in trouble? The great injured in this case is Anthropic, which was especially well considered In the business segment. But GPT-5 has improved in the programming and interface design benchmarks, two areas in which Claude, the Anthropic chatbot, used to be the preferred option for companies. According to A MENLO VENTURES STUDYClaude had so far managed Openai in the use of his models in companies. Thus, its market share was 32% for 25% of OpenAI.
Source: Menlo Ventures.
Business market dependence. Anthripic depends a lot on the use of his chatbot and his API in companies: that represents 80% of his income. The company is not going wrong, and its annual income is expected to be 17 times those obtained in 2024. In fact, in the last six months They have entered 3,000 million dollars (1 billion in June) and contracts of eight and nine figures have multiplied by three with respect to the previous year. Sources close to the company indicate that average customer expenditure has multiplied by five in the last twelve months.
GPT-5 seems to be cheaper. There is a key factor in that openai overtaking over Anthropic: GPT-5 is up to 7.5 times cheaper than Claude Opus 4.1, the latest version of that chatbot. For business clients that difference is obviously critical, and that makes them deploy it much more massively, but there is a problem in that advantage. Not for companies, of course. For OpenAi. The company took breast in June indicating that it had Three million paid business users (There were two in February), and GPT-5 seems to be promoting that figure. Anthropic Theoretically would have even more, at least according to the Menlo Ventures studio mentioned above.
But. OpenAI prices for your API indicate that GPT-5 costs $ 1.25 per million input tokens and $ 10 per million departure tokens. Claude Opus 4.1 costs 15 and $ 75 respectively at entry/output. It certainly seems that GPT-5 is much cheaper, but careful, because when GPT-5 applies reasoning, intermediate reasoning steps can be generated and those tokens are also charged even if they do not “see” in the final responses. In those cases, GPT-5 may not be “so cheap” with respect to Claude Opus 4.1, which is certainly one of the most expensive models today.
The models that “reason” succeed in companies. The success of GPT-5 in companies is due to its Improvement in programming tasks as above all in reasoning. This variant of the models has multiplied by eight their use among these clients: business clients intensively use reasoning models – more expensive, but also more precise -, especially in planning and reasoning tasks in several stages. That’s where GPT-5 seems to highlight, as they have demonstrated its results in tests such as ARC-AGI 2.
Burning money. OpenAI has created a sales team with more than 500 professionals dedicated to selling the product to companies. It is an independent effort of the alliance with Microsoft – with which they are competing in this market – and that has a problem. Those low prices are not real, because GPT-5 is still expensive to offer users. In fact, Openai is expected spend About 8,000 million dollars this year, and an even more intensive use of its models implies more infrastructure spending to provide this service.
Image | das | Fortune Brainstorm Tech
In Xataka | If the question is which of the great technology is winning the AI career, the answer is: None
GIPHY App Key not set. Please check settings