How to read this LLM pricing comparison
All prices are US dollars per million tokens at the standard tier. “Input” is the text you send, “output” is the text the model generates, and “cached input” is the discounted rate for prompt text the provider can serve from its cache. Click a column heading to sort; the default order groups models by provider from cheapest to most expensive input.
Context is the largest prompt the model accepts, where the provider lists one. Some providers charge a higher rate for prompts above 200,000 tokens; this table shows the standard rate below that size.
Where the prices come from
Every figure is read from the provider’s own pricing page rather than a third-party roundup, and the verification date is shown above the table. The sources are:
- Anthropic: Claude pricing (platform.claude.com/docs/en/about-claude/pricing)
- OpenAI: OpenAI API pricing (developers.openai.com/api/docs/pricing)
- Google: Gemini API pricing (ai.google.dev/gemini-api/docs/pricing)
Choosing on price alone
The cheapest model per token is not always the cheapest per task. A stronger model that answers correctly in one call can cost less than a weaker one that needs retries, and tokenizers differ, so the same prompt is a different number of tokens on each model. Use the LLM token counter to price your own prompt across all of them.
How to use it
- Sort by input or output price to find the cheapest models.
- Compare the cached-input column if you resend the same prompt prefix.
- Open the token counter to price your own prompt on any of these models.
Frequently asked questions
How current are the prices?
Prices were last verified on September 24, 2026 against the providers’ own pricing pages, and the date is shown next to every table. They can change without notice, so confirm a price on the provider’s page before signing a contract or setting a budget.
Why do two pricing pages show different prices for the same model?
Third-party summaries are often out of date or mix pricing tiers. This table is read from the providers’ own pages and dated, and it shows the standard tier for prompts up to 200k tokens; batch, priority and long-context rates differ.
What is the difference between input, output and cached input?
Input is the prompt you send, output is what the model generates, and cached input is prompt text the provider recognises from a recent call and bills at a reduced rate.
Are open-weight models like Llama or Mistral included?
Not yet. Open-weight models are priced by whichever company hosts them, so there is no single list price to verify. Only models with an official first-party price page are listed.
Prices verified September 24, 2026. They can change without notice; the provider’s page is the source of truth.