Thought Leadership
A report by the Financial Times describes that Anthropic’s most powerful AI model, Fable 5, is lagging behind cheaper alternatives like DeepSeek among users.
According to the report, a strategy known as model routing is becoming increasingly widespread in the industry, in which companies selectively route simple tasks to cheaper models and reserve more expensive flagship models only for particularly complex use cases. As a result, traffic and thus revenue attributable to premium offerings like Anthropic’s flagship models are declining. Between June and August 2026, Anthropic launched Claude Opus 5, a mid-priced model, for $5 per million input tokens and $25 per million output tokens. The average price paid for Anthropic models has reportedly dropped by nearly 25 percent since mid-July.
At the same time, Anthropic continues to record significant growth: The chatbot Claude reached around 245 million monthly active users in mid-2026, and the company’s annualized revenue reached around $47 billion in May. According to the report, Anthropic even overtook OpenAI in terms of enterprise spending share on AI, accounting for 41 percent of relevant corporate budgets in May. However, the growth rate is increasingly slowing as price sensitivity gains importance in enterprise procurement decisions.
Anthropic: Open models also gaining market share
Additionally, analysis firm Vercel reported that the share of open-source models in total processed token volume rose from 28 to 62 percent within two months, an indication that both OpenAI and Anthropic are increasingly losing market share to such alternatives, even as absolute usage numbers for both companies continued to rise in July. According to Anthropic’s own assessment, Opus 5 is not on par with Fable 5 when it comes to particularly high-risk, dual-use capabilities such as in cybersecurity, but according to the company, it surpasses the larger model in coding and knowledge work tasks.
(Editorial Team)
