Billing, balance, and model pricing
This page explains how MonkeyAI bills usage, how to read balance, and which models to choose at different budgets. Prices below are shown in USD / 1M tokens unless otherwise noted.
1. How billing works
- Input: everything you send to the model, including system prompts, chat history, and user messages.
- Output: the model response. This is usually more expensive than input.
- Cache read: discounted reads when upstream cache is hit.
- Image models: image generation has its own cost profile and should not be treated like text tokens.
2. Balance and top-up
- After you sign in, the dashboard shows your balance, keys, usage logs, and billing records.
- If balance is too low, requests may fail or clients may appear unresponsive.
- If a payment succeeds but balance has not refreshed yet, reload first, then check order status, then contact the admin.
3. Current pricing template
| Model | Input / 1M | Output / 1M | Cache Read / 1M | Best for |
|---|---|---|---|---|
| gpt-4o-mini | $0.25 | $2.00 | $0.018 | cost-efficient general text |
| gpt-4o | $1.75 | $14.00 | $0.175 | coding agents and long-context code work |
| gpt-4o | $2.50 | $15.00 | $0.25 | higher quality general tasks |
| gpt-4o | $5.00 | $30.00 | $0.50 | hard reasoning and premium tasks |
| claude-3-opus-latest | $5.00 | $25.00 | $0.50 | deep analysis and long writing |
| claude-sonnet-4-20250514 | $3.00 | $15.00 | $0.30 | great quality-to-cost balance |
| claude-haiku-3-5-20250520 | $1.00 | $5.00 | $0.10 | lightweight daily requests |
| deepseek-pro | $1.74 | $3.48 | $0.145 | value-focused reasoning |
| deepseek-chat | $0.14 | $0.28 | $0.028 | batch jobs and automation |
| gpt-image-2 | $75.00 | $150.00 | - | image generation |
4. Which model should I pick?
Lowest cost
deepseek-chat is good for batch jobs, automation, and budget-sensitive requests.
Best default
gpt-4o-mini is the best general low-cost recommendation for most users.
Higher quality
claude-sonnet-4-20250514 and gpt-4o are strong quality-focused choices.
Premium tasks
gpt-4o and claude-3-opus-latest fit hard reasoning, premium writing, and high-value outputs.
5. Cost control tips
- Use cheaper models first, then upgrade only when quality requires it.
- Keep prompts and conversation history compact.
- Create different API keys for different apps or projects.
- Limit output length with
max_tokenswhen possible.
6. Common billing questions
Why is my balance dropping faster than expected?
The usual cause is long context, long outputs, background polling, or a leaked API key — not the pricing page itself.
What should I do if a payment succeeded but balance did not update?
Refresh the dashboard, check order status, wait a couple minutes, then contact the admin with your account and payment timestamp.
What if my API key leaked?
Delete the old key immediately, create a new one, and update every connected client.