OVHcloud AI Endpoints: free access, limits and failover

What is free on OVH Cloud

Calls without an API key are allowed at a low rate, and OVHcloud invites you to test its models for free in a sandbox or through the API. With a key, usage is billed per model: chat LLMs are paid, and only a few non-chat models are labelled free in the catalog.

Examples listed as free when we checked: Qwen3Guard-Gen-8B , Qwen3Guard-Gen-0.6B , stable-diffusion-xl-base-v10 . Check the current catalog and your account access. Ficelle refreshes catalog data and applies eligibility checks during selection.

Listed free model examples on OVH Cloud

• Qwen3.8 27B Qwen3.8-27B

OVH Cloud free-tier limits

Anonymous calls: 2 requests per minute, per IP and per model. With an API key: 400 requests per minute, per Public Cloud project and per model. Going over returns a 429. Request bodies are limited to 2 MB (10 MB for vision models).

Official pages disagree. One OVHcloud page says an API key is required in every request, while the getting-started and capabilities pages both give an anonymous rate limit.

Get a OVH Cloud API key

In the OVHcloud Control Panel, open Public Cloud, your project, AI Endpoints, then create a key. A key requires a Public Cloud project with a payment method.

https://www.ovh.com/manager

OpenAI compatibility

LLMs support Chat Completions, Completions and the Responses API on the OpenAI-compatible base URL.

https://oai.endpoints.kepler.ai.cloud.ovh.net/v1

Data and privacy

OVHcloud says the infrastructure is in Gravelines, France, that data is not stored or shared during or after model use beyond what billing requires, and that your data is never used to train its models.

Through Ficelle, your OVH Cloud key stays on your machine and prompts go straight to OVH Cloud: there is no Ficelle proxy in between.

Before you rely on it

• Ficelle's example configuration leaves OVHcloud disabled. Its anonymous lane uses an exact model allowlist; a paid authenticated key is a different access path.
• No chat LLM is labelled free: free LLM use means anonymous calls at 2 requests per minute.
• An API key needs a Public Cloud project with a payment method.
• Models get a three-month decommission notice; several are scheduled for removal in October and November 2026.
• Catalog names and API ids can differ in format; check the exact id.

Use OVH Cloud with automatic failover

Ficelle can include OVH Cloud behind your local OpenAI-compatible endpoint when the integration is enabled and its models pass free-access and capability checks. OVH Cloud is listed in Ficelle Pro's maintained provider pack. Check your installed configuration, credentials and account entitlement before relying on it. Another eligible candidate must be available for failover to succeed. Local routing guards do not control upstream billing when an account permits paid usage beyond its free allocation. See strict-zero routing and account limits.

# OVH Cloud ships in Ficelle Pro's maintained provider pack.
ficelle doctor --text          # check this install before configuring provider access

from openai import OpenAI

client = OpenAI(base_url="http://127.0.0.1:8646/v1", api_key="ficelle-local")

# Ficelle selects among configured, enabled, eligible models.
# If none can serve the request, it returns an error.
client.chat.completions.create(
    model="ficelle/auto-tools",
    messages=[{"role": "user", "content": "..."}],
)

See what Ficelle Pro includes →

Pick a model type

Choose a model type for your request. Eligibility depends on the installed provider configuration, catalog data, free-access evidence and capability checks. A catalog listing alone does not confirm that your account can call a model.

• Tool calling
• JSON & structured output
• Fast, low-latency
• Reasoning
• Coding assistants
• Agent orchestration
• Long context
• Context compression
• Vision
• Audio input
• Video input
• Multimodal

Questions