Last verified

Prices & discounts AI glossary

What is an LLM API free tier?

A free tier is a level of LLM API access that costs nothing per token, traded for low rate limits, fewer models and, at some vendors, looser terms on how your data is used.

At a glance

  • How it works: The account is limited by requests per minute and per day rather than billed; once you exceed a limit, calls fail instead of costing money.
  • Why it matters: It is the cheapest way to prototype — but production traffic, and anything sensitive, usually belongs on a paid tier.
  • What it costs: Nothing per token; the cost is throughput, model choice and, on Google's free tier, content used to improve its products.
  • Related to: Rate limits, price per million tokens, usage limits, open-weight model.

Three different things called free

  • A free tier on a paid API: the same models, no charge, much lower limits. Google's Gemini API is the clearest example — its pricing page invites you to start building free of charge, with limited access to certain models and higher limits reserved for paid tiers.
  • A $0 model: a model whose list price is zero on both input and output, usually a small or older model a vendor offers to bring developers in. Our index counts these in the panel below.
  • Trial credits: a balance granted to a new account that is spent at normal rates. This is a discount on paid usage, not a free tier, and it runs out.

They behave differently when you scale. A free tier hits a wall at its limits, a $0 model may be withdrawn or repriced, and trial credits simply end.

In our indexModels priced at $0 in our index

$0 per token 4 of 167 current models
Vendors 2 Kwaipilot, Zhipu (Z.ai / GLM)
ModelVendorInputOutput
GLM-4.7 FlashZhipu (Z.ai / GLM)$0$0
GLM-4.5 FlashZhipu (Z.ai / GLM)$0$0
KAT-Coder-Air V1Kwaipilot$0$0
KAT-Coder-Exp 72B 1010Kwaipilot$0$0

Counted at build from our price index: models whose vendor lists both input and output at zero among the 167 current public models we price, from 33 vendors. Each rate comes from the vendor's own price page, read between 2026-07-22 and 2026-08-23. How we collect prices.

How free-tier limits work

Free access is rationed with rate limits instead of a bill. Google measures every tier against requests per minute, input tokens per minute and requests per day, and exceeding any one of them triggers an error even when the others have headroom. Its daily quotas reset at midnight Pacific time, and limits apply per project, not per API key — so splitting work across keys in one project does not add capacity.

Moving up is a billing step, not a request: Google's Tier 1 requires an active billing account linked to the project, and the upgrade from free typically takes effect immediately. Several features are not on the free tier at all — Google lists batch and flex processing as unavailable to free users.

The data question

Free tiers are not only cheaper — they can be governed by different data terms. Google's pricing page marks free-tier content as used to improve its products, and paid-tier content as not. Read the terms of the specific tier before sending customer data, source code under NDA or anything regulated through a free key. When the answer matters, the paid tier's per-token cost is the price of the stricter terms.

Using a free tier well

Free tiers suit prototypes, personal tools, coursework and evaluation runs small enough to fit the daily quota. Build with retry and backoff from the start — limit errors are the normal way a free tier says no — and keep the model name in configuration, so moving to a paid tier or a different vendor is a one-line change. Before launch, price the same workload on the paid rate card; our cost calculator does that on current rates.

Frequently asked.

Short answers to the questions that come up next.

Q · 01 Is there a free LLM API? +
Yes, in three forms: free tiers on paid APIs such as Google's Gemini API, models listed at $0 per token (counted in the panel on this page), and trial credits on new accounts. Each has limits that matter before production.
Q · 02 Does Google use free-tier Gemini API data? +
Google's pricing page marks free-tier content as used to improve its products, and paid-tier content as not. Check the current terms for the tier you use.
Q · 03 What happens when I exceed a free-tier limit? +
Requests fail with a rate-limit error until the window resets — per minute for RPM and TPM, and at midnight Pacific time for Google's daily quotas. You are not charged.

Sources · read on the date shown