AI Fervour — AI chatbots and assistant models compared | aifervour.com
Per-million-token rates look abstract until you run one draft through them — here is the arithmetic, step by step.
The listed rates: GPT-4o input at $2.50 per million tokens, GPT-4o-mini input at $0.15 per million, and output at a worked rate of $10 per million. Input and output are billed separately, and output is the expensive direction.
Step one: turn words into tokens. One English token averages 0.75 words, so multiply any word count by 1.3 — a 1,000-word draft is about 1,300 tokens.
On narrow screens, swipe or scroll the plate sideways.
Step two: apply the rate. A 2,000-word draft is about 2,700 output tokens; 2,700 × $10 ÷ 1,000,000 = $0.027 — under three cents for the generation.
Step three: scale to a heavy call. A 100,000-token input costs 100,000 × $2.50 ÷ 1,000,000 = $0.25; add a 20,000-token output at $10 per million ($0.20) and the full request is $0.45.
The same arithmetic drives the tier decision on the free-vs-paid page: at $0.45 per large call, roughly 44 calls cost as much as one $20 month. The homepage overview places these rates next to the window sizes they pay for.
Further reading