Cost Optimization
Image Tokens: pixels cost money too
Multimodal models don't bill per image — they slice images into pixel tiles and count each as Tokens. Drag the slider to change resolution and see the Token cost change in real time.
Core Billing Formula
Core Billing Formula
Token = (h̄ × w̄) / token_pixels + 2
h̄ / w̄Scaled height/width, forced to align to multiples of 32
token_pixelsPixels per Token (varies by model)
Select Model
Interactive Calculator
Width 1024 px
Height 1024 px
Presets:
Models bill based on scaled and aligned dimensions, not the original. Oversized images are shrunk; undersized are enlarged. All dimensions are finally aligned to multiples of 32.
Image pixel tile visualization
—
Token count
—
Cost vs. 512²
—
Actual billed size