Anthropic Changes Claude Code Pricing, Sparks Rivalry

Anthropic briefly removed Claude Code from the list of features on its $20 Claude Pro signup page in a small experiment, prompting immediate user outrage and social-media commentary. The company later said the change was a test affecting roughly 2% of new prosumer signups and that existing Pro and Max subscribers were unaffected. The public confusion opened an opportunity for OpenAI executives, including Sam Altman, and employees to tease the rival on social platforms as they position Codex and recent ChatGPT pricing tiers against Claude's offerings.

The visible change was limited to the marketing and signup UI, not a backend switch-off for active users, according to Anthropic statements. An Anthropic executive, Avasare, explained the experiment in product terms: engagement per subscriber has climbed substantially, usage patterns now include longer-running agentic tasks, and the original flat-rate caps were not designed for the new usage profile. OpenAI has been reacting with tiered throttles and new subscription SKUs, most recently a $100 ChatGPT Pro tier that increases Codex usage limits by 5x relative to the $20 Plus plan as a competitive counter to Claude Max.

This episode is a near-term flare in a broader industry shift: flat-rate subscriptions are under strain as agentic, long-duration workloads and heavier API usage drive marginal cost upward. The economics matter: CNBC reported a $2.5 billion run rate for Claude Code earlier this year, while OpenAI said Codex hit 3M weekly users. Providers are testing rate limits, weekly caps, and tier reshuffles rather than unilateral paywalls because of user backlash risks. The social-media exchange amplifies commercial stakes; teasing from rivals is part PR play, part competitive signaling that capacity and pricing are strategic battlegrounds.

Practical implications for practitioners Product teams and ML ops should treat this as a data point, not a final state. If you depend on hosted coding assistants for CI, developer tooling, or shipping automation, expect:

• •more tier differentiation and explicit usage meters, not unlimited flat-rate tiers;
• •tighter peak-hour caps and time-based throttles for long-running agents;
• •migration nudges from vendors designed to push heavy users into higher-margin business plans.

Vendors will disclose more formal policy changes or usage quotas in the coming weeks; monitor provider status pages, plan-change emails, and quota APIs. Also watch competitive moves: OpenAI's new $100 tier and Promos indicate aggressive capacity allocation to capture frustrated Claude users. Operationally, teams should start tracking per-session duration and agent runtimes so they can model subscription vs metered costs under tightened caps.

The immediate incident was an avoidable UX and communications error, but it exposes a structural problem: current subscription models were not built for the durability and intensity of agentic workflows. Expect iterative product changes, clearer usage metering, and public competitive theater as vendors balance unit economics against growth and retention.