Unofficial · Prices from OpenAI docs · Updated: 2026-09-05

OpenAI Batch vs Flex: Same 50% Off, Different Shape

OpenAI Batch and Flex both offer about 50% off Standard rates, but they are not the same product: Batch is an asynchronous batch API, Flex is a synchronous half-price path with higher latency. For agent evaluations and interactive flows where Batch cannot run, prefer Flex.

Flex is sync half-price. Batch is async half-price.

Both cut Standard list rates by about 50%. They are not interchangeable products. Flex is a synchronous half-price path on the same chat-style / Responses API — you send a request, wait longer, and get a completion. Batch is an asynchronous batch API — you upload a job, it runs later, you download results.

Agent evaluations do not fit Batch

If the harness is interactive — agent loops, tool calls, evals that need a completion in the same turn — Batch cannot run. Prefer Flex there. Use Batch when you can wait on a queued job and do not need live turns.

Estimate the token mix on the GPT-6 Astra pricing calculator, then confirm list rates on the LLM Pricing Leaderboard.

Not affiliated with OpenAI. List rates are unofficial estimates for planning. Always verify on the official pricing and model pages before production spend. Rates checked 2026-09-05.

Get a ping when frontier model prices change

Occasional emails when list prices move. No spam.

From the AI Cost Gate guides

Matching playbooks on the portal — budgets that auto-stop, runaway containment, and local-first cost control. Conversion stays there.