AI Model Comparison

Picking a model means trading off quality, speed, context size and price, and every provider publishes those numbers in a different place. This table puts current models from OpenAI, Anthropic, Google, DeepSeek, Mistral and Meta side by side, with prices taken from each provider’s official pricing page and the date they were last checked.

How to use the AI model comparison

1. Filter by provider with the buttons above the table, or keep All.
2. Click a column header (model, context, input or output price) to sort; click again to reverse. Models with unknown values stay at the bottom.
3. Read the notes column for what each model is designed for, prompt-length price tiers and scheduled price changes.
4. Use the API model ID shown under each name in your code, or generate a ready-made request with the prompt-to-API-code tool.

What the columns mean

Context is the context window: the maximum number of tokens the model can take in one request, counting your input and, for most APIs, the response too. A bigger window lets you send long documents or chat histories, but you still pay for every token you send. Input / 1M and Output / 1M are the standard prices in US dollars per million tokens. Output is usually several times more expensive than input. A dash means the value hasn’t been verified on the provider’s official page yet, so we show nothing rather than a guess. Vision shows whether the model accepts images as input. A question mark means the provider’s pricing page didn’t say. For open-weight models such as Llama, the price depends on who hosts the model; the note names the host used.

A table can’t tell you which model is best for your task. Use it to shortlist two or three models within your budget and context needs, then test them on real examples from your product.

Frequently asked questions

Where do the prices come from?

From each provider’s official pricing or model documentation page. Each model in our registry records the page it came from and the date it was checked, and the oldest check date is shown under the table. When two official sources disagreed, we left the value empty instead of picking one.

Why do some cells show “—”?

A dash means we couldn’t confirm the value on an official page, or the sources contradicted each other. Rather than show an outdated or invented number, we leave it blank until it can be verified.

Is a bigger context window always better?

No. It lets you send more at once, but every token still costs money and adds latency, and models can pay less attention to details buried in very long inputs. Retrieval (sending only the relevant parts) is often cheaper and more accurate than filling the window.

Why are open-weight models like Llama priced differently by different sites?

Open-weight models can be run by anyone, so the price depends on the hosting company or your own hardware. The table shows one named host as a reference; other providers may be cheaper or faster.

How often is the table updated?

Model data lives in one shared file used by all our AI tools. When providers change prices or release models we update it and the check date changes. Always confirm on the provider’s site before you commit to a budget.