Modelfold.ai has released a new token cost calculator specifically designed to estimate API spending when Claude Code calls the ModelFold endpoint. The tool, which appeared on Hacker News on November 10, 2026, clarifies a critical distinction for developers: this calculator projects pay-per-token API usage costs, not the price of a standard Claude subscription. With a current HN score of 6 and zero comments, the launch is quiet but addresses a persistent confusion in the developer community regarding billing models for AI coding assistants.

Pricing Tiers and Model Comparisons

The calculator provides detailed presets for light, regular, and heavy usage, allowing users to adjust token counts to match their actual context windows. For example, the 'Regular' preset assumes 22 days of usage with 80 requests per day, 25,000 input tokens, and 1,500 output tokens, resulting in an estimated monthly cost of $61.60 or 12,320 credits. The tool also offers a comparative view across various Claude models, showing that Claude Haiku 5.5 costs significantly less at $8.80 for the same usage profile, while Claude Opus 5.5 jumps to $114.40 and Claude Fable 5 reaches $281.60. These figures highlight the massive cost disparity between model tiers, emphasizing the need for precise budget planning when selecting a backend for Claude Code.

Integration and Billing Mechanics

Modelfold explicitly states that the calculator is not a measurement of a user's current account activity but rather a planning tool based on editable assumptions. Each request is rounded up and multiplied by the number of days and requests to derive the total cost. To utilize these prices, developers must point Claude Code at the ModelFold Anthropic-compatible endpoint using a ModelFold API key. The service charges 200 credits per dollar, and notably, failed generations are not charged, which is a crucial detail for developers worried about debugging costs during heavy iteration cycles.

Key Takeaways

  • The calculator distinguishes between ModelFold API pay-per-token costs and Anthropic's separate subscription billing.
  • Usage presets range from Light (20 days, 30 requests) to Heavy (22 days, 160 requests), with adjustable token counts.
  • Cost variance is extreme: identical usage costs $8.80 on Haiku 5.5 but $281.60 on Fable 5.
  • Failed generations are not charged, reducing financial risk during development and debugging phases.

The Bottom Line

This tool is essential for developers who need to decouple API usage costs from subscription plans. By providing transparent, editable presets and clarifying that failed requests incur no cost, Modelfold removes a major barrier to adopting Claude Code for budget-conscious projects.