From the source
Cohere is adding third-party open models to Model Vault and Model Vault Encrypted, letting enterprises run both its own models and outside open models on a fully managed, encrypted inference platform.
- The company says regulated organizations such as banks, hospitals and government departments often cannot send sensitive data to a shared API, and may lack the capacity to manage open or on-premise inference themselves.
- Model Vault runs workloads on single-tenant infrastructure in a region the customer picks, and Cohere says prompts and outputs are not retained by default.
- Model Vault Encrypted goes further, using confidential computing to keep data encrypted while the model processes it, so that neither Cohere nor any cloud provider can see prompts or outputs.
- Alongside Cohere's Command, Embed, Rerank, Parse, Transcribe and Translate models, Model Vault runs open models including
DeepSeek V4.1 Flash, GLM 5.3 Flash, MiMo V2.6 Pro, Gemma 4,
NVIDIA Nemotron, GPT-OSS and Phi.
- Model Vault is live in the United States, the European Union and Canada, while Model Vault Encrypted is live in the US with EU and Canada serving coming soon.
- Cohere says it plans more regions, regional support for Model Vault Encrypted in Canada and the EU, larger models in Model Vault Encrypted, and more models in every region.
