What AI really costs: API prices, subscription limits and GPU rent, checked daily
Measured Lab checks what AI costs to use: API prices per million tokens, what subscription plans include, how providers change their terms, and what GPU rent costs. Every number links to its source — the provider's own page or the OpenRouter listing — and carries the date we read it.
As of .
- 436 models with API prices
- 35 subscription plans
- 42 GPU offers
Compare API prices Price your task
Calculators
- Model picker Six common tasks, your volume and data rules: the models that meet them, cost a month, a cheap-first combination and what is still unknown.
- Task cost calculator Cost of one task, then per hour, week and month: growing input, cache, visible output, hidden reasoning.
- Rate limit planner A batch job and a deadline: requests and tokens per minute, the published tier that makes it, batch and rented GPUs.
- Subscription vs API Seats, active share, tasks per day and the plan’s limit window against the same work through the API.
- VRAM & GPU cost Does the model fit the card, and what a million tokens costs on your own GPU with idle time.
- Token counter Tokens for text, images and PDF pages with each provider’s counting rules.
- Invoice auditor Reads your usage export in the browser: what you paid, cheaper models of the same class, and where the money goes.
- Prompt cache auditor Paste an API request to see what breaks the provider’s prompt cache and what a fixed request saves.
Each runs in your browser on the dated prices and limits of this site. All calculators
What you can check here
- API prices per million tokensInput and output prices from the providers' own pricing pages and from OpenRouter, with history.
- Subscription limitsWhat a plan includes, in the provider's own words, with the date we read it.
- AI tool plansPlan prices of AI coding and writing tools.
Models we follow closely
- Anthropic: Claude Haiku 5.5
- Anthropic: Claude Opus 5.5
- Anthropic: Claude Sonnet 5.5
- DeepSeek: DeepSeek V4 Pro 0813
- DeepSeek: DeepSeek V4.1 Flash 0910
- Google: Gemini 3.8 Flash
- Google: Gemini 4 Argon
- Mistral: Mistral Large 4
- MoonshotAI: Kimi K3
- OpenAI: GPT-5.4 Mini
- OpenAI: GPT-6 Astra
- OpenAI: GPT-6 Sol
- Qwen: Qwen3.8 Max (0902)
- xAI: Grok 4.5
- Z.ai: GLM 5.3
What changed this week
New and removed models, price moves and terms changes, confirmed by a second check. See what changed this week
Price change briefings
Dated changes and their effect on task costs. Read the briefings
Where the numbers come from
Every number comes from a dated check of the provider's own public page, with a link to the source and the method. How we check · Report an error · About Measured Lab