Detailed Analysis
Anthropic has formalized a tiered access structure requiring users to hold at least a Pro plan ($20/month, or $17/month billed annually) in order to enable or purchase extra usage for Claude Opus and other advanced models. Extra usage functions as a pay-as-you-go overflow mechanism that activates once a subscriber exhausts their plan's included limits — approximately 45 prompts per five-hour session on Pro — and then bills at standard API rates to allow seamless continuation across Claude's primary features, including Claude Chat, Code, and Cowork. Users must navigate to Settings > Usage > Extra Usage, add a payment method, and configure spending caps or enable auto-reload, with a daily ceiling of $2,000. Free-tier users are entirely excluded from this feature, hitting hard limits with no overflow option available.
The policy effectively creates three meaningfully distinct tiers of Opus access. Pro subscribers gain baseline access to Opus alongside Sonnet and Haiku, with extra usage available as a supplement. Max 5x subscribers at $100/month receive five times the Pro usage allowance plus higher priority access and full Claude Code functionality, while Max 20x subscribers at $200/month receive twenty times the Pro baseline — a tier Anthropic positions as economically favorable for heavy users compared to equivalent API consumption costs during extended sessions. The structure reflects a deliberate strategy to monetize high-compute model access without eliminating the entry-level Pro tier, instead using the extra usage mechanism to capture marginal revenue from users whose workloads occasionally spike beyond standard limits.
The broader context situates this policy within Anthropic's ongoing effort to make frontier model access commercially sustainable. Claude Opus models are among the most computationally expensive to serve, and the company has historically struggled to balance broad accessibility with the infrastructure costs those models generate. A February 2025 promotional offer — $50 in free extra usage credit for Opus 4.6, available only to subscribers who joined before February 4 — illustrated Anthropic's willingness to use targeted incentives to drive plan adoption, though that promotion has since expired and applies to no current subscribers. Subsequent Opus releases, including Opus 4.7, operate under the same paid-plan extra usage rules, indicating the policy is now a stable part of Anthropic's commercial framework rather than a transitional measure.
This development reflects a wider industry trend in which AI companies are moving away from flat-rate unlimited access models and toward consumption-sensitive pricing that more accurately tracks the cost of serving large language models at scale. OpenAI, Google DeepMind, and others have adopted analogous structures with tiered plans and metered overflow billing. For Anthropic specifically, the move also serves a strategic function: by anchoring Opus access to paid plans and charging API-equivalent rates for overflow, the company discourages the kind of sustained high-volume free usage that would undermine the economic case for continued model development. Heavy users are thereby nudged toward Max subscriptions, which offer predictable costs at scale, while the extra usage mechanism retains Pro subscribers who might otherwise churn when hitting limits, converting those friction points into revenue rather than attrition.
Read original article →