Understanding AI Costs
ElmapiCMS uses your own API key, so all AI usage is billed directly by your provider. This page explains how AI costs work and how to keep them under control.
How AI Pricing Works
AI providers charge per token. A token is roughly 3-4 characters of English text, or about 0.75 words. Every AI request uses two types of tokens:
- Input tokens (prompt): The text you send to the AI, including system instructions and context
- Output tokens (completion): The text the AI generates in response
Providers charge different rates for input and output tokens, with output tokens typically costing 4-8x more. Pricing varies significantly between providers and models.
Provider Pricing
All prices below are per 1 million tokens (standard tier) as of July 2026.
Check each provider's pricing page for the latest rates.
OpenAI
| Model | Input | Output | Tier |
|---|---|---|---|
| gpt-5-nano | $0.05 | $0.40 | Budget |
| gpt-4.1-nano | $0.10 | $0.40 | Budget |
| gpt-4o-mini | $0.15 | $0.60 | Budget |
| gpt-5.4-nano | $0.20 | $1.25 | Budget |
| gpt-5-mini | $0.25 | $2.00 | Budget |
| gpt-4.1-mini | $0.40 | $1.60 | Budget |
| gpt-5.4-mini | $0.75 | $4.50 | Budget |
| gpt-5.6-luna | $1.00 | $6.00 | Standard |
| gpt-4.1 | $2.00 | $8.00 | Standard |
| gpt-5.4 | $2.50 | $15.00 | Standard |
| gpt-5.6-terra | $2.50 | $15.00 | Standard |
| gpt-5.5 | $5.00 | $30.00 | Premium |
| gpt-5.6-sol | $5.00 | $30.00 | Premium |
Source: platform.openai.com/docs/pricing
Anthropic
| Model | Input | Output | Tier |
|---|---|---|---|
| Claude Haiku 4.5 | $1.00 | $5.00 | Budget |
| Claude Sonnet 5 | $2.00 | $10.00 | Standard |
| Claude Sonnet 4.6 | $3.00 | $15.00 | Standard |
| Claude Sonnet 4.5 | $3.00 | $15.00 | Standard |
| Claude Opus 5 | $5.00 | $25.00 | Premium |
| Claude Opus 4.8 | $5.00 | $25.00 | Premium |
| Claude Fable 5 | $10.00 | $50.00 | Premium |
Claude Sonnet 5 has introductory pricing of $2 / $10 through August 31, 2026, then moves to the standard $3 / $15 rate.
Source: platform.claude.com/docs/en/about-claude/pricing
Google Gemini
| Model | Input | Output | Tier |
|---|---|---|---|
| Gemini 2.0 Flash-Lite | $0.075 | $0.30 | Budget |
| Gemini 2.5 Flash-Lite | $0.10 | $0.40 | Budget |
| Gemini 2.0 Flash | $0.10 | $0.40 | Budget |
| Gemini 3.1 Flash-Lite | $0.25 | $1.50 | Budget |
| Gemini 2.5 Flash | $0.30 | $2.50 | Budget |
| Gemini 3.5 Flash-Lite | $0.30 | $2.50 | Budget |
| Gemini 3 Flash Preview | $0.50 | $3.00 | Budget |
| Gemini 2.5 Pro | $1.25 | $10.00 | Standard |
| Gemini 3.6 Flash | $1.50 | $7.50 | Standard |
| Gemini 3.5 Flash | $1.50 | $9.00 | Standard |
| Gemini 3.1 Pro Preview | $2.00 | $12.00 | Premium |
Gemini 2.5 Pro and Gemini 3.1 Pro Preview charge higher rates for prompts over 200k tokens ($2.50 / $15.00 and $4.00 / $18.00 respectively).
Gemini also offers a free tier with rate limits for many Flash models, which can be useful for testing.
Source: ai.google.dev/gemini-api/docs/pricing
Google Gemini's free tier is a great way to try AI features without any cost. The rate limits are generous enough for light usage and testing.
What Uses Tokens
| Feature | Typical Usage | Notes |
|---|---|---|
| AI Assistant (chat) | Medium to high | Each message includes system prompt, conversation history, and tool call results |
| Generate content | Low to medium | Depends on prompt length and output length |
| Improve / Rewrite | Low | Sends the field text + a short instruction |
| Expand | Low to medium | Sends original text, returns expanded version |
| Summarize | Low | Sends original text, returns shorter version |
| Fix Grammar | Low | Sends original text, returns corrected version |
| Translate (field) | Low | Sends field text, returns translation |
| Translate (full entry) | Medium to high | One AI call per translatable field in the entry |
Model Tiers Explained
Budget Models
Models like gpt-5.4-nano, Claude Haiku 4.5, and Gemini 3.5 Flash-Lite cost fractions of a cent per request. They handle most content editing tasks (rewrite, translate, grammar fix, summarize) with good quality. These are the best choice for inline content tools.
Standard Models
Models like gpt-5.6-terra, Claude Sonnet 5, and Gemini 3.6 Flash provide better reasoning and follow instructions more reliably. They are recommended for the AI Assistant, which requires tool calling and multi-step planning. The cost per request is still low for typical CMS tasks.
Premium Models
Models like gpt-5.6-sol, Claude Opus 5, and Gemini 3.1 Pro offer the highest quality output. Unless you need top-tier writing or reasoning, standard models are usually sufficient for CMS work.
Cost Estimates
To give you a sense of real costs, here are estimates for common tasks using a budget model and a standard model. These assume typical content lengths (a few hundred words per field).
| Task | Budget Model | Standard Model |
|---|---|---|
| Single field rewrite or grammar fix | ~$0.0001 | ~$0.001 |
| Generate a paragraph | ~$0.0003 | ~$0.003 |
| Translate a single field | ~$0.0002 | ~$0.002 |
| Translate a full entry (10 fields) | ~$0.003 | ~$0.03 |
| AI Assistant: simple question | ~$0.001 | ~$0.01 |
| AI Assistant: create project with 3 collections | ~$0.02 | ~$0.10 |
These are rough estimates. Actual costs depend on your specific model, content length, and conversation history. Enable Show Token Usage in AI Settings to see real numbers.
Cost Factors
Model Choice
This is the biggest factor. As you can see from the pricing tables above, the difference between a budget model and a premium model can be 50-100x. Choose the cheapest model that gives you acceptable quality.
Conversation Length
The AI Assistant sends previous messages as context with each new message. Longer conversations use more tokens per message because the full history is included. The Conversation Memory setting in AI Settings controls how many messages are sent.
Content Length
Longer text fields use more tokens when processed by AI tools. A short title uses far fewer tokens than a 2,000-word article.
Full Entry Translation
Translating an entire entry makes one AI call per translatable field. An entry with 5 text fields will use roughly 5x the tokens of a single field translation.
Tips for Reducing Costs
Use a Budget Model for Content Tools
The inline AI tools (rewrite, translate, grammar) work well with smaller, cheaper models. You do not need a premium model for fixing grammar or translating a paragraph.
Keep Conversations Short
Start new conversations in the AI Assistant when switching tasks. Long conversations accumulate context that increases token usage with every message.
Lower Conversation Memory
Reduce the Conversation Memory setting in AI Settings. The default of 10 messages is a good balance, but lowering it to 6-8 can cut costs for the AI Assistant without a noticeable quality drop.
Lower Max Output Tokens
If the AI tends to generate longer responses than you need, lower the Max Output Tokens setting. This caps how much the AI can write per response.
Translate Fields Individually
If you only need a few fields translated, use the inline Translate tool on specific fields instead of translating the entire entry.
Monitor Token Usage
Enable Show Token Usage in AI Settings to see how many tokens each conversation uses. The inline content tools also show usage in the sidebar of the content form. This helps you spot unexpectedly high usage.
Checking Your Usage
Each provider has a dashboard where you can monitor your API usage and costs:
- OpenAI: platform.openai.com/usage
- Anthropic: console.anthropic.com/settings/cost
- Google Gemini: aistudio.google.com (usage section)
Most providers also let you set spending limits and alerts to avoid unexpected charges.