sference

What is sference?

sference is a managed inference platform for open models, running on European compute. It serves frontier open-weight models through an OpenAI-compatible API, with zero data retention on by default and no training on customer data.

Its pitch is precise rather than broad: in sference's own words, it is a European route, not a European label. Inference and storage of customer content run on infrastructure inside the EEA, architecturally rather than by policy, and sference publishes its DPA and its list of sub-processors instead of negotiating them one customer at a time.

sference at a glance

A European route, not a European label

"Hosted in Europe" is easy to claim and hard to verify. A model can carry an EU tag and still be served from wherever capacity happens to be free at request time.

sference takes the opposite approach. According to sference, every model available through Eden AI runs on GPU infrastructure in Finland, and its public sub-processor list names the companies and countries involved. For a legal or procurement team, that is the difference between a sales claim and a sentence that can go into a data processing agreement.

The checkpoint you test is the one that serves your traffic

sference pins every model to a specific version, and you can see it in the model name: DeepSeek-V4.1-Flash-1017, DeepSeek-V4-Flash-0731. When a provider silently swaps the weights behind a model name, an agent that passed its evaluation can start behaving differently without anyone knowing why. A pinned version turns an upgrade into a decision you make.

When should you choose sference?

Choose sference if:

• You need frontier open models, such as Kimi K3 or GLM-5.3, served from European compute.
• Zero data retention has to be the default, not an option someone must remember to enable.
• You want the model version you evaluated to stay the one in production.
• Your compliance team needs a published DPA and sub-processor list to sign off.

Consider another provider if:

• You need image, audio or video generation: sference serves language models only on Eden AI for now.
• You need a model sference does not serve yet, such as Muse Spark Open or MiniMax M3.1 Flash.
• You want the lowest possible price per token and data location does not matter for your workload.

sference pros and cons

sference models on Eden AI

Try sference through Eden AI

sference sits on Eden AI alongside the rest of the catalogue, behind one key and one bill:

import requests

response = requests.post(
    "https://api.edenai.run/v3/chat/completions",
    headers={"Authorization": "Bearer YOUR_EDENAI_API_KEY"},
    json={
        "model": "sference/moonshotai/Kimi-K3",
        "messages": [{"role": "user", "content": "Hello!"}],
    },
)
print(response.json()["choices"][0]["message"]["content"])

Because the provider is one field in the request, you can compare sference with another route for the same model on your own prompts, and switch without changing anything else in your code.

To keep every request in Europe, send the same request to Eden AI's EU endpoint, https://api.eu.edenai.run/v3, or set routing: {"zero_data_retention": true}. All five sference models qualify for both.

Frequently asked questions about sference on Eden AI