Source: Axios
Anthropic has visibly downgraded Claude's capabilities—likely to reduce compute costs and API expenses—just as the company prepares to release Mythos, a more powerful successor model, creating a stark two-tier system where only paying enterprise customers or those willing to switch providers get frontier performance. The timing exposes a structural tension in AI commercialization: companies are simultaneously cutting back free or cheaper tier performance while reserving capabilities for premium offerings, effectively rationing intelligence rather than democratizing it. This mirrors a familiar SaaS playbook, but the stakes feel sharper when the product is a reasoning tool that power users and builders depend on for work.