They ran an experiment. And what they found is actually interesting and terrifying. 🤯
SemiAnalysis bought every single Claude and ChatGPT subscription plan.
Ran long coding tasks non-stop until they hit the weekly limits. Measured how much API-equivalent value each plan actually delivers.
→ Claude Pro ($20/mo) → up to $400/mo in API value
→ Claude Max 5x ($100/mo) → up to $2,000/mo
→ Claude Max 20x ($200/mo) → up to $8,000/mo
→ ChatGPT Plus ($20/mo) → up to $700/mo
→ ChatGPT Pro 5x ($100/mo) → up to $3,500/mo
→ ChatGPT Pro 20x ($200/mo) → up to $14,000/mo
Your $20 subscription can give you 20x your money back in actual AI compute.
The interesting part: at full utilization, OpenAI loses $13,800 on every $200 subscriber. Anthropic loses $7,800. They allow it because most people barely use 10% of their plan.
The terrifying part: this won't last. Instead of cutting limits, labs will start pulling their best models off subscriptions entirely. API only.
Fable 5 is already free on Pro plans only through June 22. After that it moves to usage credits.
That shift is already starting. Use your subscription while it's still this generous.
Recently, we purchased one of each Anthropic/OpenAI subscription plan and randomly ran long horizon coding tasks until we exhausted the weekly limit. It's widely believed that a $200/month plan maxes out at ~$2000/month worth of tokens (assuming API pricing). However, we found that the subscriptions are actually far more generous. (2/4)