Anthropic and OpenAI Just Gave Us a Glimpse Into the Future of Model Pricing
Anthropic's new peak-hour throttling for Claude opens a window for OpenAI to lure users with unlimited Codex access.
Anthropic has implemented a significant policy shift for its Claude AI models, introducing dynamic throttling during peak business hours. According to a post by Anthropic technical team member Thariq Shihipar, users on all subscription tiers—including the $100/month Claude Max plan—will now consume their five-hour session limits faster on weekdays between 5am-11am PT. The company estimates this will cause about 7% of users, particularly on Pro tiers, to hit limits they previously wouldn't have. Anthropic suggests running "token-intensive background jobs" during off-peak hours, framing the change as a necessary step to manage growing demand, but it effectively penalizes users for accessing the service during standard work times.
This move has created a strategic opening for competitors. On the same day as Anthropic's announcement, OpenAI engineering lead Thibault Sottiaux revealed the removal of all usage caps for its Codex coding assistant. Sottiaux encouraged users to "build unlimited things with Codex," a clear bid to attract developers and enterprises frustrated by access restrictions elsewhere. The timing suggests a direct competitive play, with OpenAI leveraging its scale to offer uncapped access—at least temporarily—to capture market share. This back-and-forth signals a pivotal moment in commercial AI, where providers are experimenting with pricing and access models that balance massive computational costs against user expectations for reliable, on-demand service.
- Anthropic throttles Claude AI usage for all tiers during peak hours (5am-11am PT), affecting even $100/month Max subscribers.
- OpenAI removed all usage caps for its Codex coding assistant on the same day, in a likely competitive response.
- The shift highlights the infrastructure strain of scaling AI and a potential new cycle of access-based competition between providers.
Why It Matters
For businesses, reliable AI access during work hours is now a variable cost, forcing strategic decisions on which vendor to bet on.