AI Cost Calculator
Compare API prices across GPT-5, Claude, Gemini, DeepSeek & more.
Estimate what it costs to build with AI — and get a model recommendation.
Highest measured intelligence on every board (SWE-bench Verified 95%, MMMU-Pro 92.7%).
Benchmarks normalized 0–100 from public leaderboards; prices are list API rates. Verified 2026-06-29. Treat orderings as solid and exact figures as approximate; estimates are directional, not quotes.
Cost & how-to by task
DIY-with-AI steps, the right tools, and the hire cost — for each thing you want to make.
AI model API prices (2026-06-29)
USD per 1M tokens, list rates. Input is what you send; output is what the model writes back.
| Model | Provider | Tier | Input | Output |
|---|---|---|---|---|
| Claude Opus 4.8 | Anthropic | Frontier | $5 | $25 |
| Claude Sonnet 4.6 | Anthropic | Frontier | $3 | $15 |
| Claude Haiku 4.5 | Anthropic | Budget | $1 | $5 |
| Claude Fable 5 | Anthropic | Frontier | $10 | $50 |
| GPT-5.5 | OpenAI | Frontier | $5 | $30 |
| GPT-5.5 Pro | OpenAI | Reasoning | $30 | $180 |
| GPT-5.4 | OpenAI | Frontier | $2.5 | $15 |
| GPT-5.4 mini | OpenAI | Mid-range | $0.75 | $4.5 |
| GPT-5.4 nano | OpenAI | Budget | $0.2 | $1.25 |
| Gemini 3.1 Pro | Frontier | $2 | $12 | |
| Gemini 3.5 Flash | Mid-range | $1.5 | $9 | |
| DeepSeek V4 Flash | DeepSeek | Open / hosted | $0.14 | $0.28 |
| DeepSeek V4 Pro | DeepSeek | Open / hosted | $0.435 | $0.87 |
| Grok 4.3 | xAI | Frontier | $1.25 | $2.5 |
| Grok Code Fast 1 | xAI | Coding | $1 | $2 |
| Mistral Large 3 | Mistral | Open / hosted | $0.5 | $1.5 |
| Mistral Medium 3.5 | Mistral | Frontier | $1.5 | $7.5 |
| Mistral Small 4 | Mistral | Budget | $0.15 | $0.6 |
| Codestral 25.08 | Mistral | Coding | $0.3 | $0.9 |
| Sonar | Perplexity | Budget | $1 | $1 |
| Sonar Pro | Perplexity | Mid-range | $3 | $15 |
| Llama 4 Maverick | Meta · Together | Open / hosted | $0.27 | $0.85 |
| Llama 4 Scout | Meta · Together | Open / hosted | $0.18 | $0.59 |
Popular AI subscriptions
Flat monthly plans — usually the cheaper route for steady, hands-on use.
💬 Chat assistants
Higher limits, advanced models, image generation
Best for: Regular individual use
Near-unlimited, pro reasoning models, max compute
Best for: Power users
Team workspace, admin, no-training, SSO
Best for: Small teams (min 2 seats)
More usage, Claude Code, projects, research
Best for: Individuals & light coders
5× Pro usage, higher caps, early access
Best for: Heavy daily users
20× Pro usage, highest caps, priority
Best for: Power users & heavy Claude Code
Gemini 3.x Pro, higher limits, 2TB storage
Best for: Individuals
5× usage, 20TB, YouTube Premium, priority
Best for: Advanced creators/devs
20× usage, full top-tier features
Best for: Highest-end users
👨💻 Coding assistants
Unlimited completions, cloud agent, ~$15/mo AI credits
Best for: Individual devs
Premium models (incl. Opus), ~$70/mo credits, audit logs
Best for: Power-user devs
Org management, policy, ~$19/mo credits/seat
Best for: Teams (per seat)
Unlimited Tab, ~$20 frontier-model credit pool, MCP
Best for: Individual devs
~3× usage on all frontier models
Best for: Heavy daily users
~20× usage + priority new features
Best for: Max-usage power users
Credit-metered app/site generation
Best for: Building UIs/sites with AI
🔎 Research
Unlimited Pro search, model choice, file upload, Labs
Best for: Researchers
Fastest retrieval, top models, Sora video
Best for: Power researchers
🎨 Image generation
~3.3 fast GPU hours, no relax mode
Best for: Hobby image makers
15 fast hours + unlimited relax
Best for: Regular creators
30 fast hours + relax + stealth
Best for: Pros needing privacy
60 fast hours + more concurrency
Best for: Highest-volume creators
How much does AI actually cost?
There are two ways to pay for AI. Subscriptions are flat monthly plans (ChatGPT Plus, Claude Pro and Gemini AI Pro all sit at about $20/month) — best when a human is using the tool hands-on all day. The API charges per token, so you only pay for what you actually process — best when you’re automating something or building a product.
Token prices vary enormously. Budget and open models (DeepSeek, Gemini Flash-Lite, hosted Llama, GPT-5.4 nano) cost cents per million tokens; frontier reasoning models like GPT-5.5 Pro run $30–$180 per million output tokens. Output is almost always 3–5× pricier than input, and prompt caching can cut repeated-input cost by ~90% — which is why the same task can be ten times cheaper on one setup than another.
Use the calculator above to put real numbers on your case: compare every model at your monthly volume, estimate what a website or app would cost to build with AI, or get a straight model recommendation for your task and budget. All prices are list rates compiled on 2026-06-29 from official provider pricing pages.
Frequently asked questions
How much does it cost to use AI?
It depends on whether you pay per token (API) or a flat subscription. API prices range from about $0.10 per million tokens for budget models up to $30+ for top reasoning models. For everyday personal use, a $20/month subscription (ChatGPT Plus, Claude Pro, Gemini AI Pro) is usually cheaper and simpler than the API.
How much does it cost to build a website with AI?
In raw API tokens, a one-page site is typically a few dollars and a multi-page site $5–$30. But most people use a flat coding subscription (Cursor or v0 at $20/month, or Claude Code via a Max plan) which usually works out cheaper than metered API for building. Use the 'Build X with AI' tab above to estimate your case.
API vs subscription — which is cheaper?
Subscriptions win for steady human use (chatting, writing, coding all day): a flat $20 covers a lot. The API wins when you're automating — running a chatbot, batch-processing content, or building a product — because you only pay for the tokens you actually use, and caching can cut input cost by 90%.
What is a token?
A token is roughly ¾ of a word. 1,000 tokens ≈ 750 words. Both what you send (input) and what the model writes (output) are billed, and output is usually 3–5× more expensive than input.
Which AI model is cheapest?
Among capable models, open and budget tiers like DeepSeek V4 Flash, Gemini 2.5 Flash-Lite, GPT-5.4 nano and hosted Llama 4 are the cheapest per token. For frontier quality at the best value, Claude Sonnet 4.6 and Gemini 3.1 Pro lead. Use the Compare tab to rank them for your exact usage.