What does each AI coding plan really give you?
Flat-rate plans are often subsidised. We convert every plan into API-equivalent dollars, output tokens and a single score, next to the intelligence of the models you can actually use.
[ Products we track ]
Assuming you use your full plan allowance
Plans ranked by value
Click any column to sort and any plan to open its vendor page. To use your own usage numbers, open Filters.
Composite score by plan
Each plan's 0 to 100 score blends subsidy, intelligence and capacity using the weights in Filters. Hatched columns are estimates, solid columns are vendor-published.
Value multiple: API value ÷ price
How many dollars of API list-price usage each dollar of subscription buys. Hatched bars are estimates, solid bars are vendor-published. The dashed line is break-even with just paying the API.
Price vs API-equivalent value
Points above the 1x line get more than they pay for. Log scales on both axes: every step up the diagonals is a bigger subsidy.
- Published by the vendor
- Estimated (editorial)
Not scored yet
Products worth tracking where I could not verify plan pricing from a readable source.
- Z.ai GLM Coding PlanGLM-5.3 (AA 45)
Pricing page is rendered client-side and could not be read, so no plan price is scored. Models are very strong for the price per AA data.
- Kimi CodeKimi K3 (AA 44)
Plan tiers and prices are behind a client-rendered page (in Chinese); not verified.
- Zed / BYOK and Claude Code API keysany
Pay-as-you-go API use is the 1.0x baseline every plan is compared against.