Understanding AI Pricing
Tokens, tiers, input vs output — AI pricing can seem confusing. Here's everything you need to understand how much AI costs and how to get the best value.
What Are Tokens?
AI models don't read words — they read tokens. A token is roughly 3-4 characters, or about 75% of a word in English. Here's what that means in practice:
~750
words per 1,000 tokens
~4
pages per 1,000 tokens
~1.3
tokens per word (average)
So if you send a 500-word email to an AI and it writes a 300-word reply, that's roughly 670 input tokens and 400 output tokens.
Why Input and Output Are Priced Differently
Most AI APIs charge separately for input tokens (what you send) and output tokens (what the AI writes back). Output tokens are almost always more expensive because generating new text requires more computation than reading it.
Think of it like a consultant: Reading your brief (input) takes less effort than writing the deliverable (output). AI pricing reflects this same dynamic.
Currently, prices range from $0.020 per million input tokens (Mistral Nemo) to $20.00 per million input tokens (O3 Pro). That's a 1000x difference — so choosing the right model matters.
How Much Does Typical Usage Cost?
For most individual users, the costs are remarkably low. Here are some real-world examples:
| Task | ~Tokens | Cost (cheap) | Cost (premium) |
|---|---|---|---|
| Quick question & answer | 500 | < $0.001 | $0.01 |
| Email draft | 1,500 | < $0.01 | $0.03 |
| Blog post (1,000 words) | 3,000 | $0.01 | $0.06 |
| Summarise a 10-page document | 5,000 | $0.01 | $0.08 |
| Analyse a full book | 100,000 | $0.10 | $1.50 |
Consumer Subscriptions vs API Pricing
There are two ways to pay for AI:
Subscription (Consumer)
- ChatGPT Plus: $20/month
- Claude Pro: $20/month
- Gemini AI Ultra: $20/month
Fixed monthly fee, unlimited (or high-limit) usage. Best for individual users who want a simple chat interface.
API (Developer)
- Pay per token: exact usage billing
- Multiple models: choose price/quality tradeoff
- Full control: temperature, system prompts, etc.
Pay only for what you use. Best for developers building applications, or power users who want fine-grained control.
See all current API prices: LLM Comparison Table
How to Get the Best Value
1. Use the right model for the job
Don't use a $15/1M token model for simple tasks. Use a cheap model for routine work and reserve premium models for hard problems.
2. Write concise prompts
Every word in your prompt costs tokens. Be specific but not repetitive. Remove filler text.
3. Set max output limits
Tell the AI to keep responses concise. "In 3 sentences..." saves tokens compared to getting a 5-paragraph essay.
4. Check our Value Score
Our Value Score ranks models by quality-per-dollar. High value doesn't mean low quality — it means you're getting the most intelligence for your money.