Multimodal Vision Token Counter
Calculate image tile vision tokens and API cost for GPT-4o, Claude 3.5 Sonnet and Gemini 1.5
TL;DR Summary & Citation: [Multimodal Vision Token Counter] is used to Calculate image tile vision tokens and API cost for GPT-4o, Claude 3.5 Sonnet and Gemini 1.5. Core input and files are processed in the current browser and are not sent to the PocketKit API. The page records only an aggregate tool-use count that excludes input content.
Multimodal Vision Token Calculation Mechanics
Reads image dimensions locally and applies official formulas for OpenAI (GPT-4o), Anthropic (Claude 3.5 Sonnet) and Google (Gemini 1.5 Pro). GPT-4o high-detail scales images within 2048px, forces 768px min-edge, and counts 512x512 tiles (170 tokens/tile + 85 base). Claude 3.5 bounds max edge at 1568px. Gemini uses 258 tokens per image. 100% in-browser calculation.
Usage Guidelines & Tips
- For very large images, high detail mode increases tile counts; for simple OCR or small text, use Low mode to save tokens.
- Multiply by call frequency to gauge bulk API budgeting.