# sference is live on Eden AI: European inference for frontier open models
> sference's managed inference is now available through Eden AI. Kimi K3, GLM-5.3, GLM-5.3 Flash, DeepSeek V4.1 Flash and DeepSeek V4 Flash are callable with an Eden AI API key today, running in Finland with zero data retention on by default.
- Author: Jernej Štrasner (Co-founder & CEO)
- Published: 2026-10-29
- Tags: partnership, eden-ai, europe, inference
- Source: https://sference.com/blog/sference-is-live-on-eden-ai

---

sference's managed inference is now available through [Eden AI](https://www.edenai.co/). Five models are callable through the Eden AI API today: Kimi K3, GLM-5.3, GLM-5.3 Flash, DeepSeek V4.1 Flash and DeepSeek V4 Flash. They run in Finland, with zero data retention on by default.

Eden AI is a European AI gateway with one API in front of around 1,000 AI models. If your stack already goes through Eden AI, sference is now one of the providers behind it.

![sference and Eden AI partnership banner](https://sference.com/blog/sference-is-live-on-eden-ai/assets/eden_partnership.png)

## What this changes

**One key, one bill.** Teams already on Eden AI reach our models with the same API key and the same bill they use for everything else. There's no second account, contract or integration. For teams evaluating European inference, that removes most of the reason to put off the test.

**Data stays in Finland.** Inference runs on GPUs inside the EEA, in Finland for every model on Eden AI. Zero data retention is on by default, we don't train on customer data, and our [DPA is public](https://sference.com/legal/dpa).

**Pinned versions.** Every model is pinned to a specific version, and on Eden AI you can see it in the model name: `DeepSeek-V4.1-Flash-1017`, `DeepSeek-V4-Flash-0731`. The checkpoint you evaluated is the checkpoint that serves your traffic, and the serving version is recorded on every request.

**Expert models as tools.** Eden AI puts expert models next to ours: web search, OCR, translation, speech, and image and video generation. An agent built on a sference model can call them as tools, including through [Eden AI's MCP server](https://www.edenai.co/docs/v3/expert-models/mcp-server). Eden AI handles routing, fallbacks and cost tracking, and we run the inference underneath.

> "European teams keep asking us for the large open models, served from European compute, with nothing kept. sference gives them exactly that, behind the same key as the rest of their stack."
>
> — Taha Zemmouri, Co-founder and CEO, Eden AI

## Models live today

Prices are USD per 1M tokens on Eden AI, the same as sference's own list rates, and may change. The live catalogue is on the [sference provider page](https://www.edenai.co/providers/sference).

| Model | Context | Input / 1M | Output / 1M | Cached input / 1M |
| --- | --- | --- | --- | --- |
| `sference/moonshotai/Kimi-K3` | 1.0M | $3.00 | $15.00 | $0.45 |
| `sference/zai-org/GLM-5.3` | 1.0M | $1.20 | $4.20 | $0.26 |
| `sference/zai-org/GLM-5.3-Flash` | 1.0M | $0.20 | $0.60 | $0.07 |
| `sference/deepseek-ai/DeepSeek-V4.1-Flash-1017` | 1.0M | $0.50 | $1.50 | $0.05 |
| `sference/deepseek-ai/DeepSeek-V4-Flash-0731` | 1.0M | $0.28 | $0.56 | $0.07 |

Kimi K3 carries the full one-million-token context window with tool calling and adjustable reasoning effort. GLM-5.3 is a good general production default. DeepSeek V4.1 Flash 1017 is the latest DeepSeek checkpoint we serve, and [our benchmark notes](https://sference.com/blog/deepseek-v41f-1017-benchmark) cover what changed. GLM-5.3 Flash and DeepSeek V4 Flash are there for workloads where cost per token decides the architecture: classification, extraction and enrichment runs that never needed a frontier model. All five support tool calling, structured output and prompt caching. Kimi K3, GLM-5.3 Flash and DeepSeek V4.1 Flash also take image input.
## Get started

If you're already on Eden AI, switch the model string:

```javascript
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.EDENAI_API_KEY,
  baseURL: "https://api.edenai.run/v3",
});

const completion = await client.chat.completions.create({
  model: "sference/deepseek-ai/DeepSeek-V4.1-Flash-1017",
  messages: [{ role: "user", content: "Hello" }],
});
console.log(completion.choices[0].message.content);
```

Naming `sference/...` pins the request to us. If you'd rather name just the model and let Eden AI route, you can put sference first in the provider order:

```javascript
const completion = await client.chat.completions.create({
  model: "glm-5.3",
  messages: [{ role: "user", content: "Hello" }],
  routing: { order: ["sference"] }, // or { allowed_providers: ["sference"] }
});
```

For EU-only traffic, send the same request to Eden AI's EU endpoint, `https://api.eu.edenai.run/v3`, or set `routing: { zero_data_retention: true }`. sference models qualify for both.

Eden AI's chat endpoint is OpenAI-compatible. Their documentation covers [routing and fallbacks](https://www.edenai.co/docs/v3/llms/provider-routing) and the [EU endpoint](https://www.edenai.co/docs/v3/data-governance/eu-endpoint).

If you'd rather come to us directly for dedicated deployments, fine-tunes and distillations, or volume commitments, that path hasn't changed: start with the [docs](https://sference.com/docs) or [sign up](https://app.sference.com).

## About Eden AI

Eden AI is a European AI gateway: one API across around 1,000 AI models from many providers, covering language, vision, speech, documents and generative media, with an EU endpoint, routing and fallbacks, and one bill. Founded in 2020 in Lyon, France.