AI Cost Calculator
Updated Free mock REST & WebSocket APIs
Planning an AI feature? Enter your daily active users, how often each one calls the model and the tokens per call, and see the monthly LLM bill for every major model — plus cost per user, so you can price your product.
| Cached /1M | |||||
|---|---|---|---|---|---|
GPT-5 nanocheapest OpenAI In $0.05 · Out $0.40 /1M · 400K | $0.05 | $0.005 | $0.40 | 400K | $45.00 |
gpt-oss-20b OpenAI · Groq In $0.075 · Out $0.30 /1M · 128K Open-weight model; price shown is Groq's. | $0.075 | — | $0.30 | 128K | $45.00 |
Ministral 3 8B Mistral In $0.15 · Out $0.15 /1M · 256K | $0.15 | — | $0.15 | 256K | $56.25 1.3× cheapest |
Gemini 2.5 Flash-Lite Google In $0.10 · Out $0.40 /1M · 1M Gemini 2.5 models are limited to existing users. | $0.10 | $0.01 | $0.40 | 1M | $60.00 1.3× cheapest |
Claude Haiku 5.5 Anthropic In $0.10 · Out $0.50 /1M · 1M Prompts over 100,000 tokens are billed at $0.50 / $2.50 per 1M for the whole request. | $0.10 | $0.01 | $0.50 | 1M | $67.50 1.5× cheapest |
GPT-6 Luna OpenAI In $0.10 · Out $0.50 /1M · 1.05M | $0.10 | $0.01 | $0.50 | 1.05M | $67.50 1.5× cheapest |
GPT-4o mini OpenAI In $0.15 · Out $0.60 /1M · 128K | $0.15 | $0.075 | $0.60 | 128K | $90.00 2.0× cheapest |
gpt-oss-120b OpenAI · Groq In $0.15 · Out $0.60 /1M · 128K Open-weight model; price shown is Groq's. | $0.15 | — | $0.60 | 128K | $90.00 2.0× cheapest |
Mistral Small 4 Mistral In $0.15 · Out $0.60 /1M · 256K | $0.15 | — | $0.60 | 256K | $90.00 2.0× cheapest |
GPT-5.6 Luna OpenAI In $0.20 · Out $1.20 /1M · 1.05M | $0.20 | $0.02 | $1.20 | 1.05M | $150.00 3.3× cheapest |
GPT-5.4 nano OpenAI In $0.20 · Out $1.25 /1M · 400K | $0.20 | $0.02 | $1.25 | 400K | $153.75 3.4× cheapest |
Codestral Mistral In $0.30 · Out $0.90 /1M · 128K | $0.30 | — | $0.90 | 128K | $157.50 3.5× cheapest |
DeepSeek Flash DeepSeek In $0.30 · Out $1.20 /1M · 1M Peak-hour price. Off-peak is 50% off: $0.15 / $0.60. | $0.30 | $0.006 | $1.20 | 1M | $180.00 4.0× cheapest |
Gemini 3.1 Flash-Lite Google In $0.25 · Out $1.50 /1M · 1M | $0.25 | $0.025 | $1.50 | 1M | $187.50 4.2× cheapest |
GPT-5 mini OpenAI In $0.25 · Out $2 /1M · 400K | $0.25 | $0.025 | $2 | 400K | $225.00 5.0× cheapest |
GPT-4.1 mini OpenAI In $0.40 · Out $1.60 /1M · 1.05M | $0.40 | $0.10 | $1.60 | 1.05M | $240.00 5.3× cheapest |
Gemini 2.5 Flash Google In $0.30 · Out $2.50 /1M · 1M Gemini 2.5 models are limited to existing users. | $0.30 | $0.03 | $2.50 | 1M | $277.50 6.2× cheapest |
Gemini 3.5 Flash-Lite Google In $0.30 · Out $2.50 /1M · 1M | $0.30 | $0.03 | $2.50 | 1M | $277.50 6.2× cheapest |
Mistral Large 4 Mistral In $0.68 · Out $2.09 /1M · 1M Sale price; list price is $1.36 / $4.18 per 1M. | $0.68 | $0.07 | $2.09 | 1M | $360.75 8.0× cheapest |
Llama 3.3 70B Meta (Llama) · Together AI In $1.04 · Out $1.04 /1M · 128K | $1.04 | — | $1.04 | 128K | $390.00 8.7× cheapest |
Grok Build 0.1 xAI In $1 · Out $2 /1M · 256K | $1 | $0.20 | $2 | 256K | $450.00 10× cheapest |
Gemini 3.8 Flash Google In $0.75 · Out $3.75 /1M · 1M Price rises to $1.50 / $7.50 per 1M on Jan 1, 2027. Explicit cache storage is billed per hour. | $0.75 | $0.075 | $3.75 | 1M | $506.25 11× cheapest |
GPT-5.4 mini OpenAI In $0.75 · Out $4.50 /1M · 400K | $0.75 | $0.075 | $4.50 | 400K | $562.50 13× cheapest |
Grok 4.3 xAI In $1.25 · Out $2.50 /1M · 1M | $1.25 | $0.20 | $2.50 | 1M | $562.50 13× cheapest |
o4-mini OpenAI In $1.10 · Out $4.40 /1M · 200K | $1.10 | $0.275 | $4.40 | 200K | $660.00 15× cheapest |
Claude Haiku 4.5 Anthropic In $1 · Out $5 /1M · 200K | $1 | $0.10 | $5 | 200K | $675.00 15× cheapest |
DeepSeek V4 Pro DeepSeek In $1.32 · Out $3.96 /1M · 1M Peak-hour price. Off-peak (most hours and weekends) is 50% off: $0.66 / $1.98. | $1.32 | $0.044 | $3.96 | 1M | $693.00 15× cheapest |
Mistral Medium 3.5 Mistral In $1.50 · Out $7.50 /1M · 256K | $1.50 | — | $7.50 | 256K | $1,013 23× cheapest |
Grok 4.7 xAI In $2 · Out $6 /1M · 500K | $2 | $0.50 | $6 | 500K | $1,050 23× cheapest |
Gemini 2.5 Pro Google In $1.25 · Out $10 /1M · 1M Gemini 2.5 models are limited to existing users. | $1.25 | $0.125 | $10 | 1M | $1,125 25× cheapest |
GPT-5 OpenAI In $1.25 · Out $10 /1M · 400K | $1.25 | $0.125 | $10 | 400K | $1,125 25× cheapest |
GPT-4.1 OpenAI In $2 · Out $8 /1M · 1.05M | $2 | $0.50 | $8 | 1.05M | $1,200 27× cheapest |
o3 OpenAI In $2 · Out $8 /1M · 200K | $2 | $0.50 | $8 | 200K | $1,200 27× cheapest |
Claude Sonnet 5 Anthropic In $2 · Out $10 /1M · 1M | $2 | $0.20 | $10 | 1M | $1,350 30× cheapest |
Claude Sonnet 5.5 Anthropic In $2 · Out $10 /1M · 1M | $2 | $0.10 | $10 | 1M | $1,350 30× cheapest |
Gemini 4 Argon Google In $2 · Out $10 /1M · 1M Price rises to $4 / $20 per 1M on Dec 1, 2026. Caching is implicit only. | $2 | $0.10 | $10 | 1M | $1,350 30× cheapest |
GPT-6 Sol OpenAI In $2 · Out $10 /1M · 1.05M | $2 | $0.20 | $10 | 1.05M | $1,350 30× cheapest |
GPT-6.1 Sol OpenAI In $2 · Out $10 /1M · 1.05M | $2 | $0.10 | $10 | 1.05M | $1,350 30× cheapest |
Gemini 3.1 Pro (preview) Google In $2 · Out $12 /1M · 1M | $2 | $0.20 | $12 | 1M | $1,500 33× cheapest |
GPT-4o OpenAI In $2.50 · Out $10 /1M · 128K | $2.50 | $1.25 | $10 | 128K | $1,500 33× cheapest |
GPT-5.6 Terra OpenAI In $2 · Out $12 /1M · 1.05M | $2 | $0.20 | $12 | 1.05M | $1,500 33× cheapest |
GPT-5.4 OpenAI In $2.50 · Out $15 /1M · 1.05M | $2.50 | $0.25 | $15 | 1.05M | $1,875 42× cheapest |
Claude Sonnet 4.6 Anthropic In $3 · Out $15 /1M · 1M | $3 | $0.30 | $15 | 1M | $2,025 45× cheapest |
Claude Opus 5.5 Anthropic In $4 · Out $20 /1M · 1M | $4 | $0.20 | $20 | 1M | $2,700 60× cheapest |
GPT-5.6 Sol OpenAI In $4 · Out $20 /1M · 1.05M Promotional pricing through at least Nov 21, 2026. | $4 | $0.40 | $20 | 1.05M | $2,700 60× cheapest |
Claude Opus 4.8 Anthropic In $5 · Out $25 /1M · 1M | $5 | $0.50 | $25 | 1M | $3,375 75× cheapest |
Claude Opus 5 Anthropic In $5 · Out $25 /1M · 1M | $5 | $0.50 | $25 | 1M | $3,375 75× cheapest |
GPT-5.5 OpenAI In $5 · Out $30 /1M · 1.05M | $5 | $0.50 | $30 | 1.05M | $3,750 83× cheapest |
Claude Fable 5.1 Anthropic In $10 · Out $50 /1M · 1M Anthropic's most capable widely released model. | $10 | $0.25 | $50 | 1M | $6,750 150× cheapest |
GPT-6 Astra OpenAI In $10 · Out $50 /1M · 1.05M OpenAI's most capable model. | $10 | $1 | $50 | 1.05M | $6,750 150× cheapest |
Prices checked on 2026-10-18 from each provider's official pricing page (standard tier, USD per 1M tokens). Prices change often — verify before you budget.
Sources: OpenAI · Anthropic · Google · Mistral · Meta (Llama) · DeepSeek · xAI. Excludes taxes, tool/search fees, image/audio tokens and cache-write or storage surcharges.
🔒 Everything is calculated in your browser — prompts you paste are never sent anywhere.
How to estimate the AI cost of an app
- Enter daily active users and the average number of AI requests each makes per day.
- Enter typical input tokens (system prompt + history + context) and output tokens per request.
- Add your expected cache hit rate, then compare the monthly cost per model and divide by users for cost per user.
From users to dollars
requests / day = daily active users × requests per user cost / month = cost per request × requests / day × 30 cost per user = cost / month ÷ monthly active users
Example: 500 daily users × 10 requests, 2,000 input + 500 output tokens each = 5,000 requests a day. That's $225.00 / month on GPT-5 mini, $67.50 on Claude Haiku 5.5 and $1,350 on Claude Sonnet 5.5.
Typical token budgets
| Feature | Input / request | Output / request |
|---|---|---|
| Classification, tagging, routing | 300–1,000 | 5–50 |
| Chat assistant turn (with history) | 1,500–6,000 | 200–800 |
| RAG answer over retrieved docs | 5,000–40,000 | 200–600 |
| Summarizing a long document | 20,000–200,000 | 500–2,000 |
| Coding agent step | 20,000–150,000 | 500–4,000 |
Ways to cut AI costs
- Route by difficulty. Send easy requests to a small model and escalate only when needed.
- Cache the prefix. Keep system prompts and tools identical across requests to get cache hits.
- Trim history. Summarize old turns instead of resending the whole conversation.
- Batch offline work. Nightly jobs via the Batch API are half price.
- Cap output. Set
max_tokensand ask for concise formats like JSON.
Prices were checked on 2026-10-18 from each provider's official pricing page. They change often, so confirm on the provider's site before committing to a budget.
Related tools
Frequently asked questions
How much does it cost to run an AI chatbot?
A bot handling 1,000 conversations a day at 2,000 input / 500 output tokens per reply costs about $45.00 a month on GPT-5 mini and $270.00 on Claude Sonnet 5.5, before any caching.
How do I calculate cost per user?
Divide the monthly cost by monthly active users. Heavy users skew this, so also check the cost of your 95th-percentile user to set usage limits.
Does this include hosting or embeddings?
No — only LLM text tokens. Embeddings, vector databases, image/audio tokens and your own servers are extra.
Are these prices up to date?
Prices were checked on 2026-10-18 from each provider's official pricing page. They change often, so confirm on the provider's site before committing to a budget.
Your business here? Sponsor this tool from $16/week. See placements