AI for coding,
hosted in France.
Code with open-source LLMs served from our own NVIDIA GPU servers in Paris and Lyon. Connect VS Code with our free extension, or any tool that accepts an OpenAI-compatible API, and use chat and agent mode on your code: it stays in France and is never used to train models.
Open-source coding models · OpenAI-compatible API · Free VS Code extension
A coding LLM in your editor, working on your code.
AI for coding means working with a large language model (LLM) from your editor.
- What it does: explains code, writes functions and tests, edits files in agent mode.
- Quality depends on: the model and the context it gets from your project.
- With Arkane Cloud, you control: the model, the cost per request and where your code is processed (in France).
Proprietary
Arkane Cloud
Models
The vendor's own models
Open-source: Qwen, DeepSeek, GLM, Kimi…
Where code is processed
Depends on the provider
Paris and Lyon, France
Training on your code
Check the provider's terms
Never
Custom models
Set by the vendor
Deploy your own fine-tuned model
What slows down AI coding at work, and how we solve it.
Developers adopt coding assistants quickly. Security, legal and finance teams ask the same four questions first.
Model choice
The right coding LLM for each task
Switch between open-source model families such as Qwen, DeepSeek, GLM, Kimi, MiniMax and GPT-OSS from one API, and compare them on your own repositories.
Cost
Predictable cost at team scale
Pay per token on the Models API, or run a dedicated endpoint on Serverless GPU, billed per second and scaled to zero when your team is offline.
Code privacy
Your source code stays in France
Requests are processed on our own servers in Paris and Lyon. Your prompts, code and outputs are never used to train models.
No lock-in
Open models, standard API
The API is OpenAI-compatible, so tools that accept a custom base URL can use it. Change a URL and a key to switch, and leave just as easily.
Pick the right product for AI coding.
Connect VS Code to the Models API with the free extension, deploy a dedicated coding model when your team needs control, and run your own stack on GPUs for automation at volume. Same console, same account.
Per token
Models API
Call 30+ open-source models through one OpenAI-compatible endpoint, directly from VS Code chat and agent mode with our free extension.
Best for: individual developers, testing coding models, variable usage
Explore the Models API →Per second
Serverless GPU
Give your team a dedicated coding model: any catalog or vLLM-compatible Hugging Face model, including private or fine-tuned ones, with scale-to-zero.
Best for: team endpoints, custom or fine-tuned code models
Explore Serverless GPU →Dedicated
Cloud GPU
Run your own serving stack on NVIDIA GPU instances hosted in Europe, with 1 to 8 GPUs, SSH access and per-second billing, with no commitment.
Best for: CI pipelines, code review automation, full control
Explore Cloud GPU →Recommended GPUs
On Serverless GPU and reserved GPU Clusters, our own servers in Paris and Lyon: RTX PRO 6000 Blackwell for small and mid-size coding models · B300 for the largest models and long contexts
Popular coding models
Qwen, DeepSeek, GLM, Kimi, MiniMax and GPT-OSS model families, all open-source and served in France
Which coding model for which task?
A starting point for picking a model in the Models API catalog. Switch models in the extension at any time and keep the one that works best on your code.
Task
Models to try first
Why
Agent mode and multi-file changes
Kimi-K3, GLM-5.3, Deepseek-V4-Pro
Large open models, suited to planning a task and editing several files in one run
Quick questions and explanations in chat
Qwen3.6-35B-A3B, Deepseek-V4-Flash, MiniMax-M3
Lighter models with a lower price per token, suited to short, frequent requests
Code review and CI automation at volume
GPT-OSS 120B, Deepseek-V4.1-Flash, MiniMax-M2.7
Low per-token prices for large batches of diffs, or a dedicated endpoint on Serverless GPU
Suggestions to start testing, not benchmark results. Results depend on your languages and codebase. Prices and the full model list are on the pricing page.
Code with open-source LLMs in VS Code, in four steps.
Install
Add the free Arkane Cloud extension to VS Code from the Visual Studio Marketplace.
Connect
Create an API key in the Arkane Cloud console and add it to the extension.
Choose
Pick a coding model from the Models API catalog: Qwen, DeepSeek, GLM, Kimi and more.
Code
Ask questions in chat or let agent mode edit your files, with your code processed in France.
Which option for your coding team?
A quick guide, from a single developer to automated code review across a whole organization.
Workload
Recommended
Why
Individual developer in VS Code
VS Code extension + Models API
Free extension, you only pay for the tokens you use
Team with code privacy requirements
Code processed in France and never used for training, with a dedicated endpoint if needed
Other tools with a custom OpenAI base URL
OpenAI-compatible: set the base URL and your API key
Custom or fine-tuned code model
Deploy any vLLM-compatible Hugging Face model, including private ones, billed per second
CI and code review automation at volume
Per-second billing, dedicated capacity and full control over the stack
Questions about AI for coding
What does AI for coding mean?
Using a large language model to help you write, explain, review and refactor code. You call a coding LLM from your editor or your tools: in VS Code with our free extension, or from any tool that accepts an OpenAI-compatible API. In agent mode, it can plan and apply changes across several files.
What is the best LLM for coding?
There is no single best LLM for coding: results depend on your languages, your codebase and your tasks. Open-source families such as Qwen, DeepSeek, GLM, Kimi, MiniMax and GPT-OSS are widely used for code. With the Models API, you can test several of them on your own repositories and keep the one that works best.
How do I use Arkane Cloud models in VS Code?
Install the free Arkane Cloud extension from the Visual Studio Marketplace, create an API key in the console and pick a model. You can then use it in VS Code chat and agent mode. Other tools that accept a custom OpenAI-compatible base URL can use the same API key.
Is Arkane Cloud an alternative to proprietary coding assistants?
Yes. You keep your editor and switch the model behind it to open-source LLMs hosted in France, through the free VS Code extension or any tool that accepts an OpenAI-compatible base URL.
Is my source code used to train models?
No. Your prompts, code and outputs are never used to train models. Requests are processed on our own NVIDIA GPU servers in colocation datacenters in Paris and Lyon, France.
Can I use an open-source AI coding assistant without managing servers?
Yes. With the VS Code extension and the Models API, open-source coding models run on our infrastructure: there is nothing to deploy or size. Move to Serverless GPU when your team wants its own dedicated endpoint.
Can my team run its own or a fine-tuned coding model?
Yes. Deploy a catalog model or any vLLM-compatible Hugging Face model, public or private, on Serverless GPU. You get an OpenAI-compatible endpoint on the GPU of your choice, billed per second with scale-to-zero.
How is AI coding on Arkane Cloud billed?
The VS Code extension is free. Models API usage is billed per million input, output and cached tokens. Serverless GPU and Cloud GPU are billed per second of GPU time.
Other use cases
Code with open models in France, today.
Create your account, get an API key and connect the free VS Code extension. Talk to us when your team needs a dedicated endpoint.
Free playground: 10 free requests · Free VS Code extension