Crusoe Updates
Track model and hardware releases, API changes, and feature launches.
New models available: GLM-5.3 and GLM-5.3-Flash
Intelligence Foundry now supports Z.ai's GLM-5.3 and GLM-5.3-Flash models, both with a 1M-token context window. See Available models.Bug reports via API and CLI
You can now trigger, track, and download bug reports programmatically, in addition to generating them from the Console. Use the new crusoe diagnostics vm commands to collect diagnostics across many instances or from a script. See Diagnostics and the crusoe diagnostics CLI reference.Email verification codes
When you sign up with an email address, you'll now receive a six-digit verification code. Enter the code in the console to verify your email address and finish creating your account. See Create an account.Serverless Inference rate limits
Serverless Inference now enforces per-project TPM and RPM limits per model through Quotas. See Serverless rate limits.Introductory Credits
New and existing users will receive credits they can use in the Intelligence Foundry to try out Serverless Inference and Serverless Fine-Tuning—without adding a payment method. Credits burn down with usage and update in near real-time.Load Balancer Metrics
Load balancer metrics are now available in the Console. Navigate to Networking > Load Balancers, select a load balancer, then select the Metrics tab. See Load Balancer Metrics.New model available: Nemotron 3.5 Lightning 30B A3B
Intelligence Foundry now supports NVIDIA's Nemotron 3.5 Lightning 30B A3B model. See available models for Serverless Inference, Serverless Fine-Tuning, and Self-Serve Deployments.Notification Event Preferences
You can now modify your notification preferences on the Notification Settings page to focus on the alerts that matter most to you. See Notifications.Customer-managed encryption keys (CMEK)
You can now use Customer-managed encryption keys to encrypt and decrypt your data with a KMS key you own in your AWS account. See Customer-Managed Keys.AMD Bug Reports
You can now generate bug reports for AMD GPU instances directly from the Console, for both standalone VMs and Crusoe Managed Kubernetes nodes. See VM Telemetry and CMK Telemetry.Storage Metrics
VM disk I/O and NVMe drive health metrics are now available via API, Telemetry Conduit, and Crusoe MCP. See VM Telemetry.Object Storage VM Metrics
Object Storage connection latency, throughput, and retransmit metrics observed from the VM are now available via API, Telemetry Conduit, and Crusoe MCP. See VM Telemetry.You can now spin up dedicated, self-serve inference deployments on Crusoe's managed infrastructure. To get started, see Self-Serve Deployments.
Use Serverless Fine-Tuning to fine-tune an open model using a LoRA-based supervised training workflow through the Intelligence Foundry. To get started, see Serverless Fine-Tuning.
Sign-in with Github/Google
New users can now sign up with Google or Github, and existing users can link their accounts to their Google/Github to sign-in more easily in the future. Note: This doesn't apply to organization-wide enforcement of single sign-on, which is currently only supported for Okta.MI355x
AMD MI355x VMs are now available on Crusoe Cloud. Full instance specs are available on the VM Overview page. Please contact sales for availability and interest.New model available: Kimi K2.6
Intelligence Foundry now supports Kimi K2.6. See Available models.New model available: GLM 5.2
Intelligence Foundry now supports GLM 5.2. See Available models.New model available: Nemotron 3 Ultra 550B
Intelligence Foundry now supports Nemotron 3 Ultra 550B. See Available models.c2a and s2a
New AMD (Turin) CPU instance types c2a and s2a are now available in eu-norway1-a. Full instance specs are available on the VM Overview page.B200
b200-180gb-sxm-ib-8x instances are now also available in eu-norway1-a and us-west1-a, in addition to eu-iceland1-a.New model available: GLM 5.1
Intelligence Foundry now supports GLM 5.1. See Available models.New models available: DeepSeek V4 Pro, DeepSeek V4 Flash, and Nemotron 3 Nano Omni
Intelligence Foundry now supports DeepSeek V4 Pro, DeepSeek V4 Flash, and NVIDIA Nemotron 3 Nano Omni. See Available models.Object Storage
Object Storage is now generally available on Crusoe Cloud. See Object Storage.B300
B300 VMs are now available for customer POCs. Full instance specs are available on the VM Overview page. Please contact sales for availability and interest.A100
A100's are no longer available in us-west1a.New model available: Gemma 4 31B Instruct
Intelligence Foundry now supports Google's Gemma 4 31B Instruct model, the latest update to the Gemma family focused on maximizing intelligence-per-parameter. See Available models.Load Balancer Metrics
Load Balancer Metrics are now available in Limited Availability. Monitor throughput, traffic patterns, and connection behavior for your load balancers via PromQL API. See Load Balancer Metrics.Command Center now supports AMD GPU Metrics and managed Slurm Metrics.
New model available: Nemotron 3 VoiceChat
Intelligence Foundry now supports NVIDIA's Nemotron 3 VoiceChat model, a full-duplex speech-to-speech model. See Available models.Command Center is now available for Virtual Machines users. See Command Center.
Object Storage
Object Storage is released in Limited Availability on Crusoe Cloud. See Object Storage.New models available: Nemotron 3 Super, Nemotron 3 Nano, and Nemotron Cosmos
Intelligence Foundry now supports NVIDIA's Nemotron 3 Super, Nemotron 3 Nano, and Nemotron Cosmos models. See Available models.Project Roles
You can now scope user access to individual projects, rather than granting organization-wide permissions. Roles are assigned in the Console under Manage Organization > User Access > Team. See Managing Users.Crusoe MCP Server
The Crusoe Cloud Model Context Protocol (MCP) Server launched in public preview. This enables AI models and agents such as Claude and Cursor to query and analyze your Crusoe AI infrastructure. See MCP Server.Crusoe Managed Slurm
Crusoe Managed Slurm leverages Slurm's powerful job scheduling capabilities while benefiting from Crusoe Managed Kubernetes (CMK) container orchestration. See Managed Slurm.Networking Optimizations
Added guidance on applying network optimizations for shared disk migration.Billing Reservations (RIPAs)
With our new usage-based reservation system, you no longer need to manually associate individual VMs with your RIPAs. Management is now focused on monitoring your aggregate usage against your contracted capacity. See the updated docs at Reservations Overview and Managing Reservations.