Model Vault
#6 in Cloud-KI & MLOpsCohere · 2× · tolest 18. Sept. 2026
68
Momentum
Model Vault is Cohere's dedicated, fully managed SaaS Inference platform for Cohere models with isolated, single-tenant infrastructure. It offers two variants: Standard Vault (protection in transit and at rest) and the new Encrypted variant with Confidential Computing, which encrypts data during processing ("in use") via hardware-based Trusted Execution Environments (TEE) on CPU (Intel TDX/AMD SEV-SNP) and NVIDIA GPUs. The Confidential Computing feature is currently in closed Beta with limited access through Cohere sales and is free for design partners; Standard Vault is billed hourly or through commit plans per instance.
Momentum-Verloop
21.06.19.09.
Features
| Platform | Cloud SaaS, dedicated single-tenant inference, integrates with Cohere API and North platform |
| Price | Standard Vault: approx. $4-10/hour per instance ($2,500-6,500/month); Encrypted Vault: currently free for beta design partners |
| Compute Performance (FLOPS/TOPS) | No official FLOPS/TOPS figures published; sizing via performance tiers S, M, L, XL |
| Release Date | Confidential Computing in Model Vault announced September 16/17, 2026 |
| Availability | Encrypted Vault: closed beta with limited access via Cohere sales; Standard Vault: generally available (Embed/Rerank self-serve, generative models via waitlist) |
