Hi Product Hunt! Most Jev cost estimates just multiply tokens by $0.042 per million. That isn't what decides a real Jev bill, so I built a calculator that models the parts that do. Three things matter more than the per-token rate: - Fan-out: Jev evaluates many questions against one state in a single call. Asking 8 questions at once instead of in 8 calls cut the bill by 77-86% in my tests. - Rate limits: 1,200 requests/min and 250k tokens/sec. Below ~12,500 tokens per call you hit the request ceiling first. - A fixed ~260-token overhead on every request, which naive estimates leave out. To check all this I ran 2,976 real billed requests ($0.156 total). The code and raw responses are on GitHub. Billing turned out to be exactly linear, so the calculator can model it precisely. It's free, no sign-up, runs in your browser, and re-checks the published prices against TypeSafe's docs every day. It's an independent, unofficial project. What workload or assumption should it support next?

Jev Pricing

Calculate Jev costs, throughput and context limits

Calculate Jev costs, throughput and context limits