slopsubs: what a task really costs on a subscription
Effective cost vs intelligence
- API
- Claude Ultra
- ChatGPT Pro 200
- OpenCode Go
Tasks per week on each plan
How many Intelligence Index tasks a plan's weekly cap pays for when spent on one model at one effort: the cap in API-priced dollars over the task's API price. A plan's rows are alternatives, not a sum, and bundled API credit is left out.
Claude Ultra$1,846 a week at API pricesestimated cap
Theo: $200 in, $8,000 a month of API-priced inference out (as Max 20x, now Ultra). Not measured yet.
ChatGPT Pro 200$1,148 a week at API pricesestimated cap
Twice Theo's measured $573.86 a week; OpenAI doubled the allowance on 2026-10-29. Not measured yet.
OpenCode Goeach model's monthly allowance, per weekpublished cap
OpenCode's published monthly usage for each model, at list prices, with current promos, averaged over the weeks of a month
Effective cost per task, cheapest first
Each model on each selected plan, and at API prices when they are on, from low to max effort. Rows rank by their cheapest effort, so the top rows are the cheapest way to run a task at all.