Each paid plan scored out of 100. Usage per $ is how many dollars of pay-as-you-go API usage the plan's limits let you burn each month for every dollar you pay (estimated, see How We Score). Top model is the Intelligence Index of the best model the plan unlocks. Extras counts the features beyond chat: coding agents, cloud sandboxes, remote control, video, bots and bundles. Hover a row for the details, click it to open the pricing page.
No plan matches this search and these filters.
Plans with at least one model that has no cap at all: it doesn't use up your quota and doesn't stop after X requests. That's the safety net when the frontier models run out. Ranked by how smart the unlimited model is (Artificial Analysis Intelligence Index), with its speed: output tokens per second and the whole wait for a typical answer, thinking included. Not here: free models with a daily cap, and models that are unlimited only in chat (like GPT-6 Luna in ChatGPT or Flash-Lite in Gemini); you'll find those as a note on the plan's row in the value ranking.
Not free, but the quota on one model is so big that coding by hand you'll almost never hit it: hundreds of prompts or thousands of requests every 5 hours. Cheapest first, with how smart and how fast that model is (Artificial Analysis) and how big the allowance really is.
The extras that separate one $20 plan from another. A filled dot means it is in this plan, a ring means a limited or preview version. The plans follow the filters above; click a dot to see how that feature works in that plan, or a plan name to open its pricing page.
Free years, contributor programs, cheap coding plans and regional prices that beat the list price. Checked in late October 2026; offers end without notice, so read the terms before you pay.