Free LLM API credits: six providers’ free tiers compared, and what €5 actually buys

Which LLM APIs have a free tier?

The short answer, provider by provider. Conditions change often, so each row links to the page it comes from.

Free-tier terms checked on each provider’s own pages on 9 November 2026. Sources: developers.openai.com/api/docs/guides/rate-limits, ai.google.dev/gemini-api/docs/pricing, console.groq.com/docs/rate-limits, docs.together.ai/docs/billing, openrouter.ai/pricing, docs.mistral.ai/admin/billing-usage/usage-limits, ai.google.dev/gemini-api/terms, console.groq.com/docs/billing-faqs, openrouter.ai/docs/faq

Is the OpenAI API free?

Not in any way you can plan around. OpenAI’s rate-limit guide lists a Free usage tier for organisations in allowed countries, with a usage cap of $100 a month, but neither that guide nor the API pricing page mentions sign-up credits. Plan on paying from your first request; the first paid tier, Build, applies once you have bought $5 of credit.

You will find older posts promising $5 or more of free trial credit for new accounts. We could not confirm any such offer on OpenAI’s own pages today, so we do not count it.

Is the Gemini API free?

Partly. Google’s price table marks the Flash and Flash-Lite models, and the open Gemma models, as free of charge on the Free tier, with rate limits per model that you see in AI Studio. Its newest top models, such as Gemini 4 Argon and Gemini 3.1 Pro Preview, are not available on the Free tier.

Two conditions matter for a business. On the free tier, Google uses what you send and what the model returns to improve its products, and its terms say human reviewers may read it, so they ask you not to submit sensitive or personal information. And the same terms say you may use only paid services when you make your app available to users in the European Economic Area, Switzerland or the United Kingdom. The free tier is for testing, not for serving European customers.

Does Groq have a free tier?

Yes. Groq’s Free plan gives access to its hosted models with limits per model. For Qwen3.8 27B, for example, the limits are 30 requests a minute, 1,000 a day, 8,000 tokens a minute and 200,000 tokens a day; Whisper large-v3 transcription allows 2,000 requests a day and 28,800 seconds of audio a day.

That is enough to build and demo a prototype, and not enough for production traffic. To lift the limits you upgrade to the Developer tier, which asks for a payment method and bills per token.

Are OpenRouter’s free models really free?

The requests are. OpenRouter lists free variants of some models, with :free at the end of the model id, and its Free plan allows 50 requests a day across them. If you have bought at least $10 of credit, the daily cap on free models rises to 1,000 requests.

Which models have a free variant changes over time, and they run on the providers that choose to offer them. For anything beyond experiments you move to the paid model ids, where OpenRouter adds a platform fee to the credit you buy.

Do Together AI and Mistral give free API credits?

No. Together AI’s billing docs say it does not currently offer free trials, and access requires a minimum credit purchase of $15 the first time and $5 after that. The platform is prepaid.

Mistral has a free mode, but its documentation says it covers its Vibe and Studio apps and does not include API access: creating API keys needs billing activated and prepaid credit.

Rate limits or credit: which kind of free tier do you need?

The two kinds of free are good for different things. A rate-limited free tier lasts as long as the provider keeps it, but caps how much you can send. Free credit is spent at the normal rate on any model, and then it is gone.

How this works on GPU Flow

Every new account gets €5 of credit to spend on Token Factory, our API for open-source language, vision and audio models. It is for Token Factory only; it does not pay for Sandboxes or dedicated GPUs.

You activate it with a card, which is required to sign up. The card goes through a €1 verification charge that is refunded. After that, your requests are drawn from the credit at the normal price of each model, and you can spend it on any model in the catalogue, text or audio.

What you send is never used to train any model, free credit or not, and we do not store the content of your text, transcription or voice requests.

What does €5 of free credit actually buy?

It depends on the model. The table below uses current Token Factory prices and the same support answer as our pricing pages: about 500 input tokens and 300 output tokens. The last column is the number of output tokens €5 pays for if the model only wrote, which is the expensive direction.

The same credit pays for audio. This is what €5 buys with each audio model at its current price.

What happens when the free credit runs out?

Token Factory is prepaid. When your balance reaches zero, the API answers with a 402 error until you top up; nothing is charged to your card automatically. A runaway loop in your code stops at the end of your credit instead of turning into a surprise invoice.

Top-ups are in euros with VAT included, and every top-up comes with an invoice.

Related

• The cheapest LLM APIs in 2026: price per million tokens compared (OpenAI, Together, OpenRouter, GPU Flow)Price per million tokens on OpenAI, Together AI, OpenRouter and GPU Flow, each in its currency, with the cost of 1,000 requests and what makes a model cheap.
• Migrate from the OpenAI API to an open-source alternative in 15 minutes (code included)Move from the OpenAI SDK to GPU Flow in five steps: change the base URL, pick the equivalent model and keep streaming, tools and reasoning.