Support Matrix

Hardware, software, and build compatibility for Dynamo

See also: Release Artifacts for container images, wheels, Helm charts, and crates | Feature Matrix for backend feature support

At a Glance

Latest stable release: v1.2.0 — SGLang 0.5.11 (NIXL 1.0.1) | TensorRT-LLM 1.3.0rc14 (NIXL 0.10.1) | vLLM 0.20.1 (NIXL 0.10.1)

Experimental release: v1.2.0-deepseek-v4-dev.3 (DeepSeek-V4-Flash / V4-Pro on Blackwell, vLLM + SGLang containers only) — vLLM 0.20.1 | SGLang upstream deepseek-v4-blackwell preview | NIXL 0.10.1

Requirement	Supported
GPU	NVIDIA Ampere, Ada Lovelace, Hopper, Blackwell
OS	Ubuntu 22.04, Ubuntu 24.04, CentOS Stream 9 (experimental)
Arch	x86_64, ARM64 (ARM64 requires Ubuntu 24.04)
CUDA 12	Container images for SGLang and vLLM (CUDA 12.9)
CUDA 13	Container images for TensorRT-LLM (CUDA 13.1), SGLang and vLLM (CUDA 13.0)

Backend Dependencies

Driver requirements differ by backend — see CUDA and Driver Requirements below.

The following table shows the backend framework versions included with each Dynamo release:

Dynamo	SGLang	TensorRT-LLM	vLLM	NIXL
main (ToT)	`0.5.11`	`1.3.0rc17`	`0.21.0`	`0.10.1` (TRT-LLM); `1.1.0` (vLLM); `1.0.1` (SGLang)
v1.2.0	`0.5.11`	`1.3.0rc14`	`0.20.1`	`0.10.1` (TRT-LLM, vLLM); `1.0.1` (SGLang)
v1.2.0-deepseek-v4-dev.3 (experimental, partial)	upstream DSv4 preview	—	`0.20.1`	`0.10.1`
v1.2.0-deepseek-v4-dev.2 (experimental, partial)	upstream DSv4 preview	—	`0.20.0`	`0.10.1`
v1.1.1	`0.5.10.post1`	`1.3.0rc11`	`0.19.0`	`0.10.1` (TRT-LLM, vLLM); `1.0.1` (SGLang)
v1.1.0	`0.5.10.post1`	`1.3.0rc11`	`0.19.0`	`0.10.1` (TRT-LLM, vLLM); `1.0.1` (SGLang)
v1.1.0-dev.3 (experimental, partial)	`0.5.10.post1`	`1.3.0rc11`	`0.19.0`	`0.10.1`
v1.1.0-dev.2 (experimental, partial)	`0.5.9`	`1.3.0rc9`	`0.19.0`	`0.10.1`
v1.1.0-dev.1 (experimental)	`0.5.9`	`1.3.0rc5.post1`	`0.17.1`	`0.10.1`
v1.0.2	`0.5.9`	`1.3.0rc5.post1`	`0.16.0`	`0.10.1`
v1.0.1	`0.5.9`	`1.3.0rc5.post1`	`0.16.0`	`0.10.1`
v1.0.0	`0.5.9`	`1.3.0rc5.post1`	`0.16.0`	`0.10.1`
v0.9.1	`0.5.8`	`1.3.0rc3`	`0.14.1`	`0.9.0`
v0.9.0	`0.5.8`	`1.3.0rc1`	`0.14.1`	`0.9.0`
v0.8.1.post3	`0.5.6.post2`	`1.2.0rc6.post3`	`0.12.0`	`0.8.0`
v0.8.1.post2	`0.5.6.post2`	`1.2.0rc6.post2`	`0.12.0`	`0.8.0`
v0.8.1.post1	`0.5.6.post2`	`1.2.0rc6.post1`	`0.12.0`	`0.8.0`
v0.8.1	`0.5.6.post2`	`1.2.0rc6.post1`	`0.12.0`	`0.8.0`
v0.8.0	`0.5.6.post2`	`1.2.0rc6.post1`	`0.12.0`	`0.8.0`
v0.7.1	`0.5.4.post3`	`1.2.0rc3`	`0.11.0`	`0.8.0`
v0.7.0.post1	`0.5.4.post3`	`1.2.0rc3`	`0.11.0`	`0.8.0`
v0.7.0	`0.5.4.post3`	`1.2.0rc2`	`0.11.0`	`0.8.0`
v0.6.1.post1	`0.5.3.post2`	`1.1.0rc5`	`0.11.0`	`0.6.0`
v0.6.1	`0.5.3.post2`	`1.1.0rc5`	`0.11.0`	`0.6.0`
v0.6.0	`0.5.3.post2`	`1.1.0rc5`	`0.11.0`	`0.6.0`

For v1.1.0-dev.2, v1.1.0-dev.3, v1.2.0-deepseek-v4-dev.2, and v1.2.0-deepseek-v4-dev.3, the cells above match container/context.yaml on the corresponding release branch (pins used to build images). Those lines are partial releases: not every backend has a published Dynamo runtime container for that tag. See Pre-Release Artifacts for what actually shipped. The v1.2.0-deepseek-v4-dev.2 and v1.2.0-deepseek-v4-dev.3 SGLang containers are built on the upstream lmsysorg/sglang:deepseek-v4-blackwell preview image rather than a tagged SGLang release; TensorRT-LLM is not part of those dev releases.

Version Labels

1.3.0 (main / ToT) reflects the current development branch.
Releases marked (experimental, partial) are pre-releases: the table shows branch build pins, which may include backends with no NGC image for that dev tag yet.

Version Compatibility

Backend versions listed are the only versions tested and supported for each release.
TensorRT-LLM does not support Python 3.11; installation of the ai-dynamo[trtllm] wheel will fail on Python 3.11.

CUDA and Driver Requirements

Dynamo container images include CUDA toolkit libraries. The host machine must have a compatible NVIDIA GPU driver installed.

Dynamo Version	Backend	CUDA Toolkit	Min Driver	Notes
1.2.0	SGLang	12.9	575.xx+
		13.0	580.xx+
	TensorRT-LLM	13.1	580.xx+	CUDA 13 only
	vLLM	12.9	575.xx+
		13.0	580.xx+
1.1.1	SGLang	12.9	575.xx+
		13.0	580.xx+
	TensorRT-LLM	13.1	580.xx+
	vLLM	12.9	575.xx+
		13.0	580.xx+
1.1.0	SGLang	12.9	575.xx+
		13.0	580.xx+
	TensorRT-LLM	13.1	580.xx+
	vLLM	12.9	575.xx+
		13.0	580.xx+
1.0.2	SGLang	12.9	575.xx+
		13.0	580.xx+
	TensorRT-LLM	13.1	580.xx+
	vLLM	12.9	575.xx+
		13.0	580.xx+
1.0.1	SGLang	12.9	575.xx+
		13.0	580.xx+
	TensorRT-LLM	13.1	580.xx+
	vLLM	12.9	575.xx+
		13.0	580.xx+
1.0.0	SGLang	12.9	575.xx+
		13.0	580.xx+
	TensorRT-LLM	13.1	580.xx+
	vLLM	12.9	575.xx+
		13.0	580.xx+
0.9.1	SGLang	12.9	575.xx+
	TensorRT-LLM	13.0	580.xx+
	vLLM	12.9	575.xx+
0.9.0	SGLang	12.9	575.xx+
	TensorRT-LLM	13.0	580.xx+
	vLLM	12.9	575.xx+
0.8.1	SGLang	12.9	575.xx+
		13.0	580.xx+	Experimental
	TensorRT-LLM	13.0	580.xx+
	vLLM	12.9	575.xx+
		13.0	580.xx+	Experimental
0.8.0	SGLang	12.9	575.xx+
		13.0	580.xx+	Experimental
	TensorRT-LLM	13.0	580.xx+
	vLLM	12.9	575.xx+
		13.0	580.xx+	Experimental
0.7.1	SGLang	12.8	570.xx+
	TensorRT-LLM	13.0	580.xx+
	vLLM	12.9	575.xx+
0.7.0	SGLang	12.9	575.xx+
	TensorRT-LLM	13.0	580.xx+
	vLLM	12.8	570.xx+

Patch versions (e.g., v0.8.1.post1, v0.7.0.post1) have the same CUDA support as their base version.

Experimental v1.1.0-dev.* images follow the same CUDA matrix as v1.0.2. The v1.2.0-deepseek-v4-dev.3 vLLM container is CUDA 13.0 multi-arch; the SGLang containers split by arch (CUDA 12.9 on amd64, CUDA 13.0 on arm64).

Experimental CUDA 13 images are not published for all versions. Check Release Artifacts for availability.

For detailed artifact versions and NGC links (including container images, Python wheels, Helm charts, and Rust crates), see the Release Artifacts page.

CUDA Compatibility Resources

For detailed information on CUDA driver compatibility, forward compatibility, and troubleshooting:

For extended driver compatibility beyond the minimum versions listed above, consider using cuda-compat packages on the host. See Forward Compatibility for details.

Hardware Compatibility

CPU Architecture	Status
x86_64	Supported
ARM64	Supported

Dynamo provides multi-arch container images supporting both AMD64 (x86_64) and ARM64 architectures. See Release Artifacts for available images.

GPU Compatibility

If you are using a GPU, the following GPU models and architectures are supported:

GPU Architecture	Status
NVIDIA Blackwell Architecture	Supported
NVIDIA Hopper Architecture	Supported
NVIDIA Ada Lovelace Architecture	Supported
NVIDIA Ampere Architecture	Supported

Platform Architecture Compatibility

Dynamo is compatible with the following platforms:

Operating System	Version	Architecture	Status
Ubuntu	22.04	x86_64	Supported
Ubuntu	24.04	x86_64	Supported
Ubuntu	24.04	ARM64	Supported
CentOS Stream	9	x86_64	Experimental

Wheels are built using a manylinux_2_28-compatible environment and validated on CentOS Stream 9 and Ubuntu (22.04, 24.04). Compatibility with other Linux distributions is expected but not officially verified.

Cloud Service Provider Compatibility

AWS

Host Operating System	Version	Architecture	Status
Amazon Linux	2023	x86_64	Supported

AL2023 TensorRT-LLM Limitation: There is a known issue with the TensorRT-LLM framework when running the AL2023 container locally with docker run --network host ... due to a bug in mpi4py. To avoid this issue, replace the --network host flag with more precise networking configuration by mapping only the necessary ports (e.g., 4222 for nats, 2379/2380 for etcd, 8000 for frontend).

Build Support

For version-specific artifact details, installation commands, and release history, see Release Artifacts.

Dynamo currently provides build support in the following ways:

Wheels: We distribute Python wheels of Dynamo and KV Block Manager:
- ai-dynamo
- ai-dynamo-runtime
- kvbm as a standalone implementation.
Dynamo Container Images: We distribute multi-arch images (x86 & ARM64 compatible) on NGC:
- Dynamo Frontend (New in v0.8.0)
- SGLang Runtime
- SGLang Runtime (CUDA 13)
- TensorRT-LLM Runtime
- TensorRT-LLM Runtime (EFA) (New in v1.0.0, Experimental, AMD64 only)
- vLLM Runtime
- vLLM Runtime (CUDA 13)
- vLLM Runtime (EFA) (New in v1.0.0, Experimental, AMD64 only)
- Kubernetes Operator
- Snapshot Agent (New in v1.0.0, Preview)
Helm Charts: NGC hosts the helm charts supporting Kubernetes deployments of Dynamo:
- Dynamo Platform (now includes CRDs)
- Snapshot (New in v1.0.0, Preview)
- Dynamo CRDs (Deprecated in v1.0.0, CRDs managed by Operator)
- Dynamo Graph (Deprecated in v0.9.0)
Rust Crates:
- dynamo-runtime
- dynamo-llm
- dynamo-protocols
- dynamo-parsers
- dynamo-config (New in v0.8.0)
- dynamo-memory (New in v0.8.0)
- dynamo-tokens (New in v0.9.0)
- dynamo-mocker (New in v1.0.0)
- dynamo-kv-router (New in v1.0.0)

Once you’ve confirmed that your platform and architecture are compatible, you can install Dynamo by following the Local Quick Start in the README.