LLM Batch Cost & Workload Planner
Compare dated Standard and Batch text pricing, savings and request-count job estimates for OpenAI, Anthropic and Gemini.
Ready. Enter average request counts and tokens.
Batch support and limits · Reviewed 2026-10-15. USD estimates exclude media, tools, taxes and account-specific discounts. No credentials or uploads.
Enable JavaScript to compare locally.
Compare the same workload.
Request count multiplies each request’s input and output costs. Rates are selected independently for Standard and Batch; no universal discount is assumed by the calculator. Input thresholds apply to each request’s full input, including cached tokens. Mixed request sizes should be planned as separate scenarios.
Jobs are a lower bound.
For 50,001 requests and a reviewed 50,000-request limit, at least two jobs are needed by count alone. Payload size and account-specific queued-token limits may require more. No job count is inferred where a request limit is not verified. Batch is asynchronous background processing; this planner cannot predict completion time or throughput.
Cache assumptions and dated rates.
Where reviewed, cached input is an assumed warm-cache hit subset; cache creation, writes and storage are excluded. Gemini cached-input Batch comparison is unavailable because its current guide and pricing tables disagree. Rates are source-dated USD estimates, excluding tools, media, taxes and negotiated discounts. No JSONL is uploaded and no Batch jobs are submitted.