OpenAI breaks Anthropic's hegemony – and money talks.
Two weeks ago, OpenRouter users spent more money on OpenAI models than on Anthropic models.
This is a dramatic event, because it hasn't happened for over two and a half years, and the last time OpenAI led in weekly spending was on February 26, 2024.
And here's the twist: the change came almost immediately after the launch of OpenAI's new flagship model, GPT-6 Astra.
Immediately after the launch, spending on Astra surged and reached 19% of total spending on OpenAI and Anthropic models on OpenRouter.
Astra has already surpassed key Anthropic models and has become one of the most significant models in terms of financial expenditure on the platform.
This number is especially important because it's not just about usage volume.
It's about the money customers are actually willing to pay for computing power.
The data indicates that the increase in spending on Astra is mainly due to increased demand for the new model.
In other words, users didn't just try Astra to see 'what it can do.' Some of the usage actually shifted to it, and concurrently, there was a relative decrease in the share of spending on Anthropic's models.
If this trend continues, it's a significant indication that OpenAI is succeeding in taking a share of the market that Anthropic held for a long time.
In the world of AI, it's easy to get excited about test scores, programming capabilities, or impressive demonstrations.
But in the end, there's one metric that can't be faked – where customers spend their money.
The OpenRouter platform allows users to choose from a variety of models, so a change in spending can reflect a real shift between providers.
The fact that OpenAI managed to surpass Anthropic in weekly spending for the first time since February 2024 makes this data much more interesting.
The implication is broader than just one model.
The artificial intelligence market is maturing.
Users are starting to choose not only based on who presents the most impressive technology, but based on a combination of performance, price, speed, capabilities, and the work environment.
If OpenAI succeeds in maintaining Astra's momentum, it may re-strengthen its position in the API market and pressure Anthropic to continue investing even more in models, infrastructure, and pricing.
And remember, Astra's advantage comes from training based on a broader infrastructure, as reports indicate it's the first model in the world whose training cost exceeded $1 billion!
It is also the first model trained on over one hundred thousand GPUs.
Which means that computing power is an advantage for the end customer.
And the data from OpenRouter tells a very simple story:
The money started moving after this advantage!
And after more than two and a half years in which Anthropic held the advantage in weekly spending, OpenAI finally managed to reverse the trend with stronger infrastructure!
The battle between the two giants is no longer just about who built the smarter model – it's about who can train models with broader infrastructure...
And who do you call when you need to train models with broader infrastructure?
Jensen Huang.