Recurring free allocation
A no-charge allowance that resets on a published schedule. Models or regions may still be restricted.
Compare free AI APIs without confusing free tiers, trial credits, and zero-price models
Start with six provider documents reviewed on August 24, 2026. Then explore the larger imported catalog as leads to verify—not as a promise that every model or quota is current.
What the providers currently publish
These summaries point to primary documentation. Check the linked page and your account dashboard before choosing a production dependency.
| Provider | Offer type | What the primary source establishes | Source |
|---|---|---|---|
| Google Gemini API | Free tier for selected models | Limits depend on model, project, and usage tier; AI Studio shows the limits that apply to your project. | Rate limits |
| GroqCloud | Free plan | Limits are model-specific and apply at the organization level; the account Limits page is the exact source for an organization. | Rate limits |
| OpenRouter | Zero-price :free routes |
The published total is 50 free-model requests per day without at least $10 in purchased credits, and 1,000 per day after that threshold. | FAQ and limits |
| Cloudflare Workers AI | Recurring free allocation | The published allocation is 10,000 Neurons per day at no charge; some models require the Workers Paid plan. | Pricing |
| Hugging Face Inference Providers | Monthly account credit | Free users receive $0.10 in monthly credits for routed inference; the provider explicitly marks that amount as subject to change. | Pricing and billing |
| Cerebras Inference | Free tier | The free-tier table publishes model-specific request and token limits, while the account Limits section remains authoritative for an organization. | Rate limits |
Verification boundary: “Reviewed” means the linked primary source was checked on the date above. We did not create six fresh accounts or run an authenticated end-to-end API test for every offer.
Four labels that should not be merged
A useful free-model comparison identifies the economic mechanism instead of calling every offer “free.”
Zero-price model route
A model endpoint priced at zero while available. Request, token, capacity, or data-policy limits can still apply.
Trial credit
A one-time or time-limited balance. It is not a permanent free tier.
Limit not recorded
The imported dataset lacks a limit value. It never means unlimited use.
Catalog warning: The searchable directory below is imported from a shared CSV and is not independently verified row by row. Its timestamp records when the snapshot was generated, not when the claims were validated. Model names and quota notes—some of which are in Chinese—may be stale.
Need one OpenAI-compatible base URL instead of separate provider accounts?
Primordial AI is a separate gateway, not any provider's free tier. Review current models and pricing first, then create an account and a dedicated API key if the fit is right.
Free model catalog
The site loads the newest imported snapshot first and falls back to a static copy. Individual rows are not independently verified.
Free AI API questions
Is there a free LLM API?
Yes. Several providers publish a free tier, recurring free allocation, trial credit, or zero-price model route. Those offers are not equivalent, and limits can change.
Does a free model mean unlimited API usage?
No. A zero-price model can still have request, token, account, region, capacity, or availability limits.
Is the imported catalog verified row by row?
No. Its generation time is not a verification time. Use the six-provider table above or the provider's current documentation before relying on a limit.
Are these third-party offers part of Primordial AI?
No. They are separate from Primordial AI accounts, pricing, and API access.