The IPO filing for Anthropic is expected to become public soon. However, a packed question-and-answer session with OpenAI's Chief Financial Officer Sarah Friar took place on Tuesday at the annual Communacopia + Technology conference hosted by Goldman Sachs, held at the Palace Hotel in San Francisco. In her remarks, Friar made it clear that pricing has become a decisive factor in the popularity of large language models. This presents a potential challenge for Anthropic, as the market generally perceives that Anthropic's customers bear higher usage costs.
Friar noted that after OpenAI reduced the price of its GPT-5.6 Luna model, call volume for that model surged tenfold, helping OpenAI capture the highest market share on OpenRouter, an aggregation platform developers use to access multiple large models. As previously reported, on the OpenRouter platform, Luna's token call volume has already surpassed the combined total of Anthropic's Opus 5 and Sonnet 5 models by more than double. Based on Friar's comments on Tuesday, the growth in call volume has been sufficient to offset the per-unit revenue loss from the price cut.
From Anthropic's perspective, OpenAI's push to capture market share, particularly through price reductions, comes at a highly unfavorable time. One of Anthropic's proudest metrics is its annualized revenue scale, which significantly leads OpenAI. However, a price war could rewrite that dynamic. Anthropic has already responded: the company officially stated that its newly released Fable 5.1 model last week carries a price 25% lower than the previous-generation Fable 5 under standard business workloads.
There is no denying that Anthropic possesses strong technical capabilities, and some enterprise clients are willing to pay a premium for its products. Yet, as the capabilities of various large models continue to evolve, combined with the abundance of low-cost options from open-source AI, Anthropic's pricing power may be weakened. To be sure, there will always be a segment of customers seeking the most cutting-edge model performance, but the size of that customer base is likely to shrink steadily. Under such industry conditions, how can Anthropic defend its revenue advantage? This is a question that investors participating in its upcoming IPO must consider.
In related news, Meta CEO Mark Zuckerberg is highly optimistic about personal AI agents. He stated in July that such AI could "improve your life, health, relationships, finances, almost everything." Meta officially launched the Muse application on Tuesday, now available on the Apple App Store and Google Play Store. The key question now is how many everyday users will genuinely embrace the service. To promote Muse, Meta's AI executive Alexander Wang posted on the X platform, showcasing multiple "real users" sharing their experiences, though some cases seemed rather far-fetched. One user said Muse helped him discover free time in his schedule to go for a run. Does anyone really need AI to tell them when they can squeeze in half an hour of free time each day?
Zuckerberg may be overly optimistic. With widespread societal resistance to data center construction, the general public's sentiment toward AI is not particularly favorable. Convincing ordinary people to entrust their health, relationships, and finances entirely to AI sounds like a difficult proposition. In the short term, it may only be journalists who will use it for testing purposes.