Skip to main content

openai/gpt-5.4/chat

Give GPT-5.4 the rows and formulas behind a warehouse report. It can trace a surprising percentage back to the unit used in the denominator.

Input $2.50Output $15per 1 million tokensCached input $0.25Prompts over 272K tokensInput $5Cached input $0.50Output $22.50

Workbench

These examples were written for this endpoint. Replace the prompt and sample media URLs with your own accessible inputs, then check the current quote and model availability before submitting.

  • Model reply

    The mismatch is **items vs. orders**: **D2** is returned **items**, but **B2** is shipped **orders**. For **item return rate**, use shipped **items** as the denominator: **Correct formula:** `=D2/C2` **Calculation:** `=6/120 = 0.05` **Percentage:** **5%** `=D2/B2` would be returns per order, not item return rate.

    Check a return-rate formula
    • 384
    • none
  • Model reply

    If the blank February is treated as 0, the three-month average is: \[ (120 + 0 + 150) / 3 = 90 \] If you average only the reported months, it is: \[ (120 + 150) / 2 = 135 \] So, treating the blank as zero lowers the average from 135 to 90, which may understate performance if the blank means “missing” rather than “no sales.”

    Handle a missing month
    • 384
    • none
  • Model reply

    Site A’s return rate is 2/10 = 20%. Site B’s is 9/90 = 10%. Combined return rate: - Total returns = 2 + 9 = 11 - Total shipments = 10 + 90 = 100 - Combined rate = 11/100 = 11% Averaging 20% and 10% gives 15%, which is different because it weights both sites equally. The correct combined rate must weight by shipment counts, and Site B has far more shipments.

    Calculate a combined rate
    • 384
    • none

This example starts at Input $2.50 / Output $15 per 1 million tokens. The Playground updates the estimate when you change the settings.