AI
Set the workload and see what the same agent costs on 361 models, ranked, including the cache hit rate most calculators forget. Prices refresh from live catalogues every few hours.
Sonnet 5 ranks #226 of 289 models that fit. The cheapest fit, Ling 3.0 Flash, would cost $0.892 a month (99% less).
List prices, calculated in your browser. Real bills vary with batching, context growth across a conversation and provider discounts, so treat this as the baseline for comparison.
Agents resend most of their context every turn — the system prompt, the tools, the whole conversation so far. Providers price that repeated prefix at roughly a tenth of fresh input, so an agent whose prompts hit the cache 60% of the time costs less than half of what the naive calculation suggests. That is also why agent platforms live or die on context discipline: compressing history and delegating to sub-agents shows up directly in this number.
Pocket users keep this cost flat another way: bring your own API key and pay providers directly at list price — the same math this calculator shows, with no platform markup on tokens. The full LLM pricing ledger tracks where these numbers come from and when they last moved.