OVHcloud AI Endpoints

European hosting. Seven models at €0 on the catalog; the rest you can still call without an account at 2 requests a minute.

Get API key Official docs ↗

OVHcloud AI Endpoints is European serverless inference, hosted in France. The public catalog at ovhcloud.com/en/public-cloud/ai-endpoints/catalog/ lists 19 live models. Seven of them are priced at €0: NVIDIA Riva TTS (EN/DE/ES/IT), Qwen3Guard-Gen 0.6B and 8B, and Stable Diffusion XL.

Llama 3.3 70B, GPT-OSS 20B/120B and Qwen 3.5/3.6 are pay-as-you-go with a key, but you can call any of them with no account at 2 requests a minute. A registered Public Cloud project jumps to 400 a minute and starts billing. OpenAI-compatible at oai.endpoints.kepler.ai.cloud.ovh.net/v1. They say prompts are not used to train.

Free models: 7
Max context: 262K
Free tier: 7 models at €0
Requirement: Nothing — no account

Endpoint

Base URLhttps://oai.endpoints.kepler.ai.cloud.ovh.net/v1

Env varOVH_AI_ENDPOINTS_ACCESS_TOKEN

Rate limit2 requests/minute anonymous; 400 requests/minute with a key

OpenAI-compatible
text
image
audio
speech
code

Code examples

npm install openai

import OpenAI from 'openai'

const client = new OpenAI({
  baseURL: 'https://oai.endpoints.kepler.ai.cloud.ovh.net/v1',
  apiKey: process.env.OVH_AI_ENDPOINTS_ACCESS_TOKEN
})

const response = await client.chat.completions.create({
  model: 'Meta-Llama-3_3-70B-Instruct',
  messages: [{ role: 'user', content: 'Explain closures in one paragraph.' }]
})

console.log(response.choices[0].message.content)

pip install openai

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://oai.endpoints.kepler.ai.cloud.ovh.net/v1",
    api_key=os.environ["OVH_AI_ENDPOINTS_ACCESS_TOKEN"],
)

response = client.chat.completions.create(
    model="Meta-Llama-3_3-70B-Instruct",
    messages=[{"role": "user", "content": "Explain closures in one paragraph."}],
)

print(response.choices[0].message.content)

curl https://oai.endpoints.kepler.ai.cloud.ovh.net/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OVH_AI_ENDPOINTS_ACCESS_TOKEN" \
  -d '{
    "model": "Meta-Llama-3_3-70B-Instruct",
    "messages": [{"role": "user", "content": "Explain closures in one paragraph."}]
  }'

Worth knowing

The catalog prices seven models at €0: four NVIDIA Riva TTS voices, Qwen3Guard-Gen 0.6B and 8B, and Stable Diffusion XL.
Chat models (Llama 3.3 70B, GPT-OSS, Qwen 3.5/3.6) are pay-as-you-go once you have a key. Without one you can still call them at 2 requests/minute.
Five catalog ids are listed as unavailable: Mistral 7B, Mistral Nemo, Mistral Small 3.2, Qwen3-32B and Qwen3-Coder-30B.

Models you get for free7

Riva TTS Englishnvr-tts-en-us
Riva TTS Germannvr-tts-de-de
Riva TTS Spanishnvr-tts-es-es
Riva TTS Italiannvr-tts-it-it
Qwen 3Guard Gen 0.6Bqwen-guard-gen-06b
Qwen 3Guard Gen 8Bqwen-guard-gen-8b
Stable Diffusion XLstable-diffusion-xl