Products · Production-Ready

Dynamia Product Overview

Compare HAMi Enterprise and HAMi AI Platform, then choose the right path for production-grade heterogeneous compute.

Same DNA

HAMi & Dynamia

HAMi is the CNCF Sandbox project for heterogeneous AI computing virtualization. Founded and core-maintained by the dynamia.ai team. Our enterprise editions ship the same upstream code — hardened, supported, SLA-backed.

CNCF Sandbox
Open governance · vendor-neutral foundation
Created by Dynamia
HAMi maintainers are the dynamia.ai team
100% upstream
No fork drift · no lock-in
OEM-grade SLA
24/7 support · long-term release lines

Product Lineup

Choose a product to view its introduction, downloads, install guide and compatibility matrix.

2 products available

Which Product Is Right for You?

Three product lines — open-source HAMi, HAMi Enterprise, and HAMi AI Platform.

Choose in 30 Seconds

I want to experiment with GPU sharing on my own

Open-source HAMi

I need a supported, hardened GPU virtualization layer

HAMi Enterprise

I also need a multi-cluster GUI, quotas, and observability

HAMi AI Platform

I want to manage a platform for multiple internal teams

HAMi AI Platform
Capability
Open-source HAMi
HAMi Enterprise
HAMi AI Platform
GPU Virtualization Core
Open-source HAMi base — extended in commercial editions.
GPU memory sharing
GPU compute sharing
Memory hard isolation
Compute hard isolation
Memory & compute oversubscription
High-perf mode (no virtualization overhead)
HAMi Turbo mode
Auto-scaling without restart
Task priority & preemption
Idle-time analytics & memory profiling
Optimized scheduler engine (parallel)
Binpack / Spread strategies
NVIDIA topology-aware affinity
Volcano deep integration (preemption / vGPU joint)
Heterogeneous GPU Vendor Support
Broad vendor coverage from NVIDIA to leading domestic accelerators and AMD.
NVIDIA (full lineup)
AMD MI series (incl. MI300X)
Ascend (910B / 910C)
MetaX (MXC500)
Kunlunxin (P800)
Hygon DCU (K100)
Iluvatar (BI)
Cambricon (MLU370)
Enflame (SG2042)
Moore Threads (MTT S3000)
Multi-cluster Control Plane
Enterprise-only — turn-key UI, observability, lifecycle management.
Multi-cluster observability (cluster / app / GPU)
GPU hardware monitoring
Rich alert policies
Turn-key GUI (multi-cluster dashboard, quota, workload)
Visual Binpack / Spread strategy editor
Global time-range search & filters
Multi-cluster lifecycle management (LCM)
Enterprise Operations
Enterprise-only — tenant, security, billing, OpenAPI, SLA.
Enterprise-grade tenant quota
Network security policies
Tenant-level security isolation
Metering & billing
OpenAPI & automation
SLA-backed support (24/7)
Air-gap / offline deployment
Capabilities covered

Checkmark = supported in that edition

5
24
38

Need Help Choosing the Right Product?

Sizing, eval licenses, air-gap support — talk to our team.