Home / How we compare AI models
How we compare AI models
What we compare, where every number comes from, how often we check it, and how we choose our picks.
Prices checked
What we compare
25 AI models from 14 labs in the US, China and Europe, sold to developers through an API, and the 8 most-used AI chat apps, sold as monthly plans. For each model we track the release date, context window and price per million input and output tokens. For each app we track its plans and the main features built in.
Where the numbers come from
Every price is checked by hand against the company’s own pricing page or documentation. API prices are standard pay-as-you-go rates in US dollars, before batch or caching discounts. For Chinese labs we use international (non-mainland) prices. Where a price depends on prompt length or time of day, we show the standard rate and note the exception.
How often we update
All prices were last checked on November 18, 2026. We recheck every price at least monthly, and straight away when a company announces a change, such as the Gemini 4 Argon and Carbon price rise on December 1, 2026. Each page shows the date its prices were checked.
How we choose picks
The recommendations in “Pick for me” and on comparison pages are our own judgment, based on price and the features each plan includes. No company pays to be listed or recommended, and the site has no affiliate links.
Independence
Compare AI Bots isn’t affiliated with OpenAI, Anthropic, Google, xAI, Meta, Mistral, Perplexity, DeepSeek, Alibaba, Moonshot, Zhipu, MiniMax, ByteDance, Xiaomi or Tencent. Product names are trademarks of their owners.