Prompt Assay
OpenClaudeOpenChatGPT
§.FAQ

How much does a single call cost?

Typical cost ranges for the most common models and an example calculation.

Updated 2026-04-13 · By Jon Lasley

Cost is determined by tokens in × input rate + tokens out × output rate. Here's a rough guide for a medium-length prompt call (say, 2,000 tokens in, 500 tokens out):

ModelRough cost per call
Claude Opus 5$0.0225
Claude Sonnet 5$0.009
Claude Fable 5$0.045
Claude Sonnet 4.6$0.0135
Claude Haiku 4.5$0.0045
GPT-5.6 Sol$0.018
GPT-5.6 Terra$0.010
GPT-5.6 Luna$0.001
GPT-4.1 Mini$0.0016
Gemini 3.7 Flash$0.0034
Gemini 2.5 Pro$0.0075
Gemini 2.5 Flash-Lite$0.0004
Back-of-envelope. Real costs depend on your exact token counts and model tokenizer quirks. On reasoning models the thinking tokens bill as output, so a call at high or max effort can cost several times the figure above · see the effort multipliers in the Reasoning Effort help article. Gemini 3.7 Flash and 3.6 Flash are on promotional pricing that doubles on 2027-01-01.
Prompt Assay doesn't take a cut
Every call bills your provider account directly. We don't mark up tokens and we don't meter your usage for billing. The usage dashboard shows the cost at your provider's published rate.
Related
← All docsOpen the workbench →