OpenAI Flex Pricing Calculator — Standard vs Flex vs Priority

OpenAI's synchronous API lets you pick a processing lane with the service_tier request field. The lane you choose changes latency, reliability, and — the part this tool computes — the per-token price. There are three lanes with distinct prices: Standard (default), Flex (flex), and Priority (priority). A fourth value, auto, lets OpenAI decide and has no distinct price, so it is out of scope here.

Every tier uses the same token-cost formula, with input and output priced separately because generating a token costs the model more compute than reading one:

1. Convert tokens to millions: inM = inputTokens ÷ 1,000,000 and outM = outputTokens ÷ 1,000,000.
2. Standard = (inM × stdIn + outM × stdOut) × requests, using the model's published Standard input and output prices per 1M tokens.
3. Flex = Standard × 0.5. Flex tokens are billed at OpenAI's Batch-API rate, which is exactly 50% of the Standard price for every supported model — so Flex is derived from Standard, never stored separately, and the 50% invariant can never drift.
4. Priority = (inM × prioIn + outM × prioOut) × requests, using the model's published Priority prices. Priority is nota flat multiple of Standard — it ranges from roughly 2× to 2.5× depending on the model — so each model carries its own Priority price. Where OpenAI has not published one (some nano and pro tiers), Priority is shown as "not published" rather than guessed.
5. Savings for a tier = Standard − tier, and the percentage divides that by the Standard total (guarded so a zero bill shows 0%, never a divide-by-zero).

The tool cross-checks Flex two independent ways — halving the Standard total, and applying the discounted per-million Flex rates directly to the tokens — and confirms they agree to the cent, the same "verified two ways" approach the tax calculator uses against the IRD's formula. Flex's trade-off is real: it runs slower, on a best-effort basis, and can return a 429 when capacity is unavailable, in which case you are not billed. Everything is computed in your browser from stored, dated prices — there are no network calls and no data leaves the page.

• OpenAI — Flex processing guide (Batch-rate pricing, 429 on unavailable)
• OpenAI — Priority processing guide (per-token premium for low latency)
• OpenAI — API pricing (Standard / Flex / Priority per-token rates)
• OpenAI — service_tier request parameter reference
• Central Bank of Sri Lanka — indicative USD→LKR rate

Standard and Priority per-token prices were last cross-checked against OpenAI's pricing page on 2026-07-13. Pricing is reviewed quarterly and after any OpenAI tier change. The Flex = 50% of Standard rule is a published, vendor-guaranteed term.