Many people want to jump on the AI bandwagon and learn how to use AI tools, but the monthly subscription fees can still feel a bit steep for some. However, compared to APIs, it’s actually quite cost-effective.
Recently, a foreign company published a research report showing that even the most expensive plans at $200 per month—Claude Max and ChatGPT Pro—when converted to equivalent API token pricing, Claude Max at 20x is approximately equivalent to $8,000 worth of tokens, while ChatGPT Pro at 20x can reach about $14,000.
Claude Pro and ChatGPT Plus work the same way, allowing you to use several times more tokens compared to the API.
Image source: SemiAnalysis
Claude Max and ChatGPT Pro cost $200 per month, but for heavy users the equivalent API value could be tens of times higher.
Recently, SemiAnalysis released an interesting report where they actually purchased all subscription plans from Anthropic and OpenAI, then continuously used them for coding tasks until hitting the weekly limits, and finally used API pricing to back-calculate “how much it would theoretically cost if the same usage went through the API instead.”
The end result is that Claude Max 20x costs $200/month and could generate up to approximately $8,000 in API token value. ChatGPT Pro 20x at the same $200/month could be worth up to approximately $14,000 in API token value.
This also means that heavy users who max out their allowance almost every month are getting an incredible deal.
Of course, the same goes for Claude Pro and ChatGPT Plus, the two most popular subscriptions.
Claude Pro costs $20/month, converting to up to approximately $400 worth of API tokens. ChatGPT Plus costs $20/month, converting to up to approximately $700 worth of API tokens.
Overall, ChatGPT users are making more money than Claude users.
Image source: SemiAnalysis
And it doesn’t matter if you don’t use it all up. Using the highest-tier Claude Max 20x and ChatGPT Pro plans as examples, as long as you use approximately 10% and 5.7% of the quota respectively, the API-equivalent value consumed will already be close to the monthly subscription cost.
For Claude Pro users, the monthly fee matches the API token cost at just 20% usage. For ChatGPT Plus, it’s 11.4% of usage.
However, conversely, from Anthropic and OpenAI’s perspective, the more users rely on it, the bigger their losses.
Anthropic’s highest-tier plan would see its gross margin drop to -900% if users max out their weekly usage. OpenAI’s highest-tier plan would drop even further to -1650%. That’s truly insane.
Image source: SemiAnalysis
A note here: these figures are estimates calculated by SemiAnalysis using specific test workloads and API pricing, not official data from OpenAI or Anthropic. Actual costs are still influenced by many factors, including which model is used, input/output ratio, cache hit rate, long context, inference mode, system scheduling, and whether the platform uses different models for internal traffic routing.
However, this does explain why the cost of using AI has become an increasingly pressing concern for businesses.
As we enter the era of coding agents and long-running automated workflows, token consumption will accelerate dramatically. AI doesn’t just read your question—it also reads project files, context, tool return results, and finally generates output.
To solve this problem, Tom’s Hardware reports that enterprises have started turning to cheaper Chinese LLMs or open-source models, assigning routine tasks to lower-cost models while reserving high-end models like OpenAI and Anthropic for more complex tasks that genuinely require top-tier capabilities.
Source: KOCPC Chinese