Cost formula
Realtime cost = tokens / 1M × realtime price; Batch cost = tokens / 1M × batch price; Savings = realtime cost − batch cost
Calculator
Batch API vs realtime cost saver
See how much batch pricing can cut from the same input/output workload when latency is flexible.
Batch vs realtime
Batch saves 50.0%
$250.00saved vs realtime
50% batch discount on GPT-4o · worth waiting if delay is OK
Cloud realtime$500.00
Batch$250.00
Realtime / req$0.01
Batch / req$0.005
Discount50%
Workload
Request shape
Same tokens; only the delivery path changes.
Batch APIs usually finish within hours. Use realtime when latency matters; batch when overnight delay is OK and you want the 50% list discount.
Sensitivity
Savings across volumes
Same token mix at a few request counts.
Batch APIs usually trade latency for price. Updated 2026-07-31. Approximate static list prices for planning only. Always verify against each provider’s official pricing page before production budgeting.
Guide
How to get value from this calculator
Quantify savings from routing eligible workloads to batch API pricing instead of realtime endpoints. Batch discounts of up to 50% are common, but jobs run asynchronously with latency measured in hours. Use this to decide which traffic can tolerate delay in exchange for lower monthly spend.
Why it matters
Batch pricing is one of the largest straightforward discounts available, yet teams leave money on the table by sending offline jobs through realtime endpoints. Identifying even 30% of traffic as batch-eligible can materially reduce monthly API bills.
How to use it
- Enter your monthly input and output token volume for the workload under review.
- Select the model and note both realtime and batch list prices.
- Review side-by-side monthly cost for 100% realtime vs 100% batch.
- Compare dual cost scales for batch vs realtime, then check volume sensitivity at your split.
- Set latency tolerance so only workloads that can wait hours count as batch-eligible.
- Apply a realistic batch-eligible percentage (not all traffic qualifies).
- Calculate blended monthly cost based on your batch vs realtime split.
- Confirm batch API availability and discount for your model on the provider's pricing page.
Planning tips
- Nightly evals, report generation, bulk classification, and backfills are ideal batch candidates.
- User-facing chat and low-latency features must stay on realtime—do not batch those.
- Batch jobs may have separate rate limits and queue times—plan SLA accordingly.
- Input/output price asymmetry still applies in batch mode; output-heavy jobs save more in absolute dollars.
- Watch how dual scales and volume sensitivity shift when you change the batch-eligible percentage.
- Verify whether cached input pricing applies to batch requests on your provider.
Frequently asked questions
How large is the typical batch discount?
Many providers offer roughly 50% off standard input and output token rates for batch jobs, but discounts vary by model and change over time. CentsPerToken uses approximate list prices—verify officially.
Can I mix batch and realtime in one product?
Yes. Most production systems use realtime for interactive paths and batch for background processing. Model the split rather than assuming all-or-nothing.
Does batch pricing affect output token rates?
Usually both input and output rates are discounted, but confirm per model. Output-heavy workloads still pay the output premium, just at the reduced batch rate.
Are these savings guaranteed?
No. Figures are planning estimates based on static list prices. Your actual invoice depends on eligibility, successful job completion, and current provider rates.