Plan your spend
What the workload costs
A sticker price is a rate, not a bill. Pick the shape of your traffic and see what each model would actually cost per month at its cheapest listed provider — including the cache writes, reasoning tokens and long-context tiers a headline $/M figure leaves out.
Short turns, light history, little caching — an assistant in a product.
- input
- 1.5k tok
- output
- 500 tok
- reasoning
- —
- cache hit
- 20%
- context
- 8k tok
Each estimate uses the model's cheapest listed provider and bills every line the workload touches: uncached input, cache reads, cache writes, reasoning tokens and output. Where a provider publishes no cache or reasoning rate, those tokens are billed at the full input or output rate — we never assume a discount nobody offers.
Sticker prices are shown for reference only — they are not what the workload costs.