Describe the workload

Task examples measure the source or deliverable itself. Agent reasoning, tool calls, and repeated file reads can add usage, so those remain editable.

Defined as five 1,000-word source documents plus a 250-word brief, producing a 1,000-word paper. OpenAI estimates one token at about ¾ of an English word, giving about 7.0K input and 1.33K output; output is rounded to 1.4K. Source.
Processing speed
Fast: ~1.5× speed, 2× token price.
K tokens
Prompt, files, sources, history, and tool results.
K tokens
Reused input billed at the cached rate.
K tokens
The paper, analysis, code, or answer you receive.
K tokens
Optional budget for hidden reasoning. No fixed level multiplier is published.
Reasoning level
Medium is OpenAI’s balanced default for most tasks.
No automatic reasoning multiplier. Research finds that useful test-time compute depends on prompt difficulty, and excessively long reasoning can sometimes hurt performance. Enter a reasoning allowance only when you have measured usage or want a planning buffer.
Long context: off — 7.0K inputKnown-token estimate
Pricing, assumptions, and sources
ModelNew inputCached inputOutput + reasoningCurrent estimate
GPT-5.6 Luna$1.00$0.1000$6.00$0.0154
GPT-5.6 Terra$2.50$0.2500$15.00$0.0385
GPT-5.6 Sol$5.00$0.5000$30.00$0.0770

Long context: when total input is greater than 272K tokens, GPT-5.6 applies 2× input pricing and 1.5× output pricing to the full request. Maximum input is 922K, maximum output is 128K, and the total context window is 1.05M tokens.

Fast: OpenAI documents approximately 1.5× model speed. GPT-5.6 API Priority pricing is 2× Standard. ChatGPT/Codex Fast is 2.5× credits for GPT-5.6—not 2×. The calculator applies the correct multiplier for the selected billing method.

Token rates and source measurements last verified July 18, 2026. Exchange rates update once per business day via Frankfurter. Estimates exclude separate tool-call or container charges where applicable.

Calculator inputs remain in your browser and are not transmitted. Daily exchange rates are fetched from Frankfurter.