How to Use a Free AI API Without an API Key
Updated 27 October 2026.
In short: Two services answer AI requests with no account and no API key. Kilo AI Gateway has free models, up to 200 requests an hour from one IP address. OVHcloud AI Endpoints runs open models such as gpt-oss-120b, Llama 3.3 70B, Qwen and Mistral, up to 2 requests a minute per model. Both use OpenAI's chat format, so most apps and OpenAI's libraries work with them.
The two services that need no key
| Service | Free models (checked 27 October 2026) | Limit without a key | Base URL |
|---|---|---|---|
| Kilo AI Gateway | kilo-auto/free, nvidia/nemotron-3-ultra-550b-a55b:free, reflection/beam-free, poolside/laguna-s-2.1:free, cohere/north-mini-code:free and more | 200 requests an hour per IP address | https://ai-gateway.kilo.ai/api/v1 |
| OVHcloud AI Endpoints | gpt-oss-120b, gpt-oss-20b, Meta-Llama-3_3-70B-Instruct, Qwen3.8-27B, Qwen3.5-397B-A17B, Mistral-Small-3.2-24B-Instruct-2506, Mistral-Nemo-Instruct-2407 | 2 requests a minute per model, per IP address | https://oai.endpoints.kepler.ai.cloud.ovh.net/v1 |
Every model named above answered a request with no key on 27 October 2026. OVHcloud serves its models from data centres in the EU. Its docs give the anonymous limit; with a paid OVHcloud key the limit is 400 requests a minute, billed per token.
Send a request with no key
Open a command window. On Windows, press the Start key, type PowerShell and open it. On a Mac, open Terminal.
Kilo AI Gateway, Windows (PowerShell):
$body = '{"model":"kilo-auto/free","messages":[{"role":"user","content":"Explain how AI works in a few words"}]}'
$r = Invoke-RestMethod -Method Post -Uri "https://ai-gateway.kilo.ai/api/v1/chat/completions" -ContentType "application/json" -Body $body
$r.choices[0].message.content
OVHcloud, Mac or Linux (Terminal):
curl https://oai.endpoints.kepler.ai.cloud.ovh.net/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-oss-120b",
"messages": [{"role": "user", "content": "Explain how AI works in a few words"}]
}'
The current model lists are at https://api.kilo.ai/api/gateway/models and
https://oai.endpoints.kepler.ai.cloud.ovh.net/v1/models; both open without a key.
Use it in an app or in Python
Apps made for OpenAI usually ask for an API key even when the service does not need one. Enter any text as the key, the base URL from the table, and a model name. OpenAI's Python library works the same way:
from openai import OpenAI
client = OpenAI(base_url="https://oai.endpoints.kepler.ai.cloud.ovh.net/v1", api_key="no-key")
reply = client.chat.completions.create(
model="Meta-Llama-3_3-70B-Instruct",
messages=[{"role": "user", "content": "Explain how AI works in a few words"}],
)
print(reply.choices[0].message.content)
Free AI in apps, with no key
The OpenCode coding app includes its own free models, such as DeepSeek V4.1 Flash, with nothing to sign up for; they work only inside OpenCode. See how to install OpenCode.
When something goes wrong
| What you see | What it means | What to do |
|---|---|---|
429 too many requests | Over the limit for your IP address | Wait a minute (OVHcloud) or an hour (Kilo), or change the model |
404 model not found | The model name is wrong or the model was removed | Copy the exact name from the live model list |
404 "Provider returned error" (Kilo) | That free model's provider is not answering now | Change the model, for example to kilo-auto/free |
| Empty answer | A reasoning model used the whole reply length on thinking | Allow a longer reply (max_tokens) or change the model |
Words to know
- API key: the password most AI services require. See what is an API key.
- Anonymous use: calling an API with no account and no key; the limit is counted per IP address.
- IP address: the internet address your requests come from; everyone on one network often shares it.
- Base URL: the address an app sends its requests to.
Free AI API keys
For higher limits, many services give free keys with a daily allowance: see how to get AI API keys for free, which lists each one with its limit and source.
Frequently asked questions
Is there a free AI API with no API key?+
Yes. Kilo AI Gateway (free models, 200 requests an hour per IP address) and OVHcloud AI Endpoints (open models, 2 requests a minute per model) answer with no account and no key.
Which models work without a key on OVHcloud?+
On 27 October 2026: gpt-oss-120b, gpt-oss-20b, Meta-Llama-3_3-70B-Instruct, Qwen3.8-27B, Qwen3.5-397B-A17B, Mistral-Small-3.2-24B-Instruct-2506 and Mistral-Nemo-Instruct-2407.
What do I enter as the API key in an app?+
Any text. Apps made for OpenAI often require something in the key field; Kilo and OVHcloud ignore it for free use.
Why do I get error 429?+
The limit is counted per IP address: 200 requests an hour on Kilo, 2 requests a minute per model on OVHcloud. Wait, or change the model.
More free API key tutorials
- How to get a Gemini API key for free
- How to get a Groq API key for free
- How to get a Cloudflare Workers AI API key for free
- How to get a Cohere API key for free
- How to get an OpenRouter API key for free
- How to get an Experimental Labs API key for free (with the $1 verification)
- How to get a Requesty API key for free
- How to use Kilo Code free models (no API key)
- How to get an LLM7 API key for free
- How to get an Ollama API key for free
- How to get a Gonka AI API key for free
- How to get a Perplexity API key for free
- Where Qwen is free
- Where Mistral models are free
- The Grok API, checked on a new account
- Where DeepSeek models are free
- Where GLM models are free
- Where OpenAI models are free
- Every free AI API and its daily limit
By model: DeepSeek · OpenAI (GPT) · Claude · Grok · Perplexity · Mistral · Kimi · Qwen · GLM · MiniMax · Llama · Nemotron · Gemma · gpt-oss
Want to use these APIs in a desktop app? OpenCode is a free, open-source AI coding app for Windows, Mac and Linux. How to add free API keys to OpenCode and pick a model.