Is it controlled by token price or token amount?
Do i get more usage from using auto vs cursor 1.5.
Hey, good question. Here’s how it works:
How usage is calculated: it’s based on token cost (your dollar budget), not the number of requests or raw token counts. On Pro+, you have two separate pools: “Auto + Composer” and “API” ((70) USD).
Auto vs Cursor 1.5: on individual plans, both draw from the same “Auto + Composer” pool, so there’s no budget difference between them. Auto has fixed rates ((1,25) USD per 1M input, (6,00) USD per 1M output, (0,25) USD per 1M cache read) no matter which model it picks under the hood. Composer 1.5 on individual plans is also billed from the Auto pool, so the effective per-token price is lower than if you manually picked models via the API pool.
Bottom line: both Auto and Composer 1.5 are more cost-effective than choosing specific models (Claude, GPT, etc.), because those come out of your API budget.
More details here:
I guess im still confused as composer is one of the more expensive models
| Composer 1.5 | $3.5 input | - | $0.35 | $17.5 output |
|---|
To say its more cost effective than choosing specific models is wrong, unless you are saying that in the auto pool we get composer at auto pricing.
I just don’t feel like i get 3x the usage of auto over composer.
Hey, fair point. Let me clarify my previous reply.
Here’s what the docs say exactly:
- Both Auto and Composer 1.5 on individual plans draw from the same pool, “Auto + Composer”. This pool is much larger than the API pool (USD 70 on Pro+).
- Auto has fixed pricing: USD 1.25 input, USD 6.00 output, USD 0.25 cache read.
- For Composer 1.5, the docs say: “Your effective per-token cost is lower than the listed price” (USD 3.5/USD 17.5). On Team and Enterprise plans, the API price applies directly.
Either way, if you want to maximize usage, Auto is the most cost-effective option since it has the lowest per-token prices in the pool.
Looks like this didn’t age well lol.. now composer 2.0 looks to be cheaper then Auto.