How do I estimate GPT-4o token costs before I run a batch?

Quick math: tokens ≈ characters / 4 for English – but use a real tokenizer when it matters.

In Python:

Code:

import tiktoken
enc = tiktoken.encoding_for_model('gpt-4o')
tokens = len(enc.encode(your_text))

Then: input price per 1M tokens + output price per 1M tokens × your ratio. Don't forget:
- Output tokens cost more than input tokens
- Batch/async APIs are ~50% cheaper
- Long context fills up fast – trim before you send

Our LLM Token & Cost Estimator on tools.php does the whole thing live (tokens + $ estimate for GPT-4o / Claude / Gemini).

What's the biggest token surprise you've hit – runaway prompts or huge system messages?