The world's leading open models, fully under your control

Sovereignty in AI means customers can choose, inspect, and deploy the models that best serve their needs without surrendering control to a single vendor, but that freedom should never come at the expense of security or privacy. Adding leading open models from Cohere and other providers across the globe inside of Model Vault gives our customers the flexibility they need to use the right tool for the job without sacrificing the standards that govern their business.

Every organization wants to use the best AI models available for the task at hand but many cannot. For a bank, a hospital or a government department, sending sensitive data to a shared API is often off the table. At the same time, managing inference for open or on-premise models may also not be possible due to capability or capacity constraints. Teams end up choosing between the model they want, the control they need, and the capacity they have to manage them.

They shouldn't have to. Today we're launching third party models inside Model Vault and Model Vault Encrypted, allowing Enterprises to run both Cohere’s models and leading open models on a fully managed and encrypted inference platform. Zero data retention of prompts and outputs is implemented by default in Model Vault.

Two levels of protection

Model Vault provides the full inference stack managed through an isolated instance dedicated to your organization alone. Model Vault Encrypted goes further, keeping your data encrypted even while the model processes it.

Model Vault: Your models and workloads run on single tenant infrastructure. You choose the region, and your prompts and outputs stay there. Cohere manages the infrastructure so you get production-grade serving without building and running your own cluster.

Model Vault Encrypted: For your most sensitive work, Model Vault Encrypted protects data in use, not just at rest and in transit. Prompts and outputs are processed inside isolated environments and stay encrypted throughout using confidential computing. Not even Cohere can see what you send to a model or what it sends back. Neither can any cloud provider.

What sovereign inference means

Sovereign inference means you decide how your AI runs, not your provider. It comes down to four things:

• Where your data lives. Run inference workloads in your preferred supported region to help meet your organization’s data residency and regulatory requirements.
• Who can access it. Model Vault keeps your workloads in a separate, isolated tenant. With Model Vault Encrypted, no one but you can see your prompts or outputs, including Cohere.
• Which models you can use. The model that best fits the task, not just the one that happens to be available through a single provider.

What customers are saying

“Security is at the heart of what Language Weaver offers to our customers, whether that pertains to zero data retention, data residency requirements, or simply having more control over their AI model. Our customers around the globe trust us with some of their most sensitive content. Cohere Model Vault lets Language Weaver run models - including the LW Pro model - securely and efficiently, while ensuring our customers are in control over where their data goes. It’s fast and reliable, and because Cohere handles the hosting and scaling, our teams can stay focused on what we do best: delivering for our customers.”

Dragos Munteanu, VP Research and Product Development, RWS

The right model for every job

No single model is best at everything. Model Vault lets you pick the right one for each task, and run every one of them with the same protections.

Cohere's models. Command for generative and agentic work, Embed and Rerank for search and retrieval, Parse for documents, Transcribe for speech and Translate for multilingual workflows.

High performing open models. Alongside our own models, Model Vault runs the most-used open models, including DeepSeek V4.1 Flash (the most-used open model on OpenRouter), GLM 5.3 Flash, MiMo V2.6 Pro, Gemma 4, and NVIDIA Nemotron. GPT-OSS and Phi models are available too.

With Model Vault you can choose the best model for each task and run it within a unified, production inference stack.

Explore Model Vault

Where it's available

Model Vault is live today with regional serving available in the United States, the European Union and Canada. Model Vault Encrypted is live in the US, with regional EU and Canada serving coming soon.

This is the start. Over the coming months we'll bring sovereign inference to more places and more workloads:

• New regions, so more organizations can run models in their preferred region.
• Model Vault Encrypted with regional supported for Canada and the EU.
• Larger models in Model Vault Encrypted, extending confidential compute to more demanding workloads.
• More models in every region, as new open models are released and adopted.

FAQ

Which option is right for me? Model Vault suits most regulated workloads that need an isolated tenant and in-region processing with a wide model catalogue. Choose Model Vault Encrypted for data sensitive workloads, when no one outside your organization, including Cohere, should be able to see your data.

Can I run open models in Model Vault Encrypted? Yes. Models like Gemma, GPT-OSS and Phi models run in Model Vault Encrypted today, alongside Cohere's Embed, Rerank, Transcribe and Translate. See the table above for the full list.

How is Model Vault priced? Pricing is by GPU type. Please visit cohere.com/pricing for details including on-demand (hourly) pricing and promotions. Reach out to sales for reserved capacity; monthly and annual commit pricing is available.

Get started

Model Vault and Model Vault Encrypted are available today. Sign up, or talk to our team about the right setup for your organization.