Release Notes#
Release 1.6.0#
New Supported Models#
wan-ai/wan2.2 (14B text-to-video and 14B image-to-video variants)
New Features#
The OpenAI-compatible video generation endpoint,
/v1/videos/generationsis available with minor modifications for the wan-ai/wan2.2 NIM. For more details, refer to the OpenAI-compatible video generation API.Wan2.2 ships three precisions per variant (BF16, FP8, NVFP4). Choose one at container startup by setting
NIM_MODEL_PRECISION. The container validates the requested precision against the host GPU’s compute capability (CC) and fails fast before downloading any model weights if the GPU cannot run it. Refer to Precision Selection for the supported combinations.
Release 1.5.3#
New Features#
microsoft/trellis NIM now supports ARM machines, including DGX Spark.
Security updates#
black-forest-labs/flux.2-klein-4b
qwen/qwen-image
microsoft/trellis
Release 1.5.2#
New Supported Models#
nvidia/qwen-image-edit-nvpcb-ovsl2sl — an NVIDIA fine-tuned version of Qwen-Image-Edit specialized for Omniverse-to-NVPCB solder-light style transfer, enabling sim-to-real data augmentation for PCB inspection workflows. Supported by the qwen/qwen-image-edit NIM.
Release 1.5.1#
New Features#
OpenAI API-compatible endpoints are available for the flux.1 and stable-diffusion-3.5-large NIMs.
Security updates#
black-forest-labs/flux.1-dev
black-forest-labs/flux.1-kontext-dev
black-forest-labs/flux.1-schnell
stabilityai/stable-diffusion-3.5-large
Release 1.5.0#
New Supported Models#
qwen/qwen-image
qwen/qwen-image-edit
Security updates#
black-forest-labs/flux.2-klein-4b
Release 1.4.1#
Summary#
stabilityai/stable-diffusion-3.5-large supports deployment on DGX Spark and GH200 with FP8 TensorRT engine profiles.
Security updates#
black-forest-labs/flux.1-dev
black-forest-labs/flux.1-kontext-dev
black-forest-labs/flux.1-schnell
stabilityai/stable-diffusion-3.5-large
microsoft/trellis
Release 1.4.0#
New Supported Models#
black-forest-labs/flux.2-klein-4b
New Features#
OpenAI API Compatible endpoints are available for black-forest-labs/flux.2-klein-4b NIM
Release 1.3.4#
New Features#
microsoft/trellis now supports multi-image input for improved output quality.
Security updates#
black-forest-labs/flux.1-dev
black-forest-labs/flux.1-kontext-dev
black-forest-labs/flux.1-schnell
stabilityai/stable-diffusion-3.5-large
Release 1.3.3#
Summary#
black-forest-labs/flux.1-dev supports deployment on DGX Spark and other Arm-based systems. Includes CVE fixes for black-forest-labs/flux.1-schnell and black-forest-labs/flux.1-kontext-dev.
Release 1.3.2#
Summary#
black-forest-labs/flux.1-schnell and black-forest-labs/flux.1-kontext-dev support deployment on DGX Spark and other Arm-based systems.
Release 1.3.1#
Summary#
New model profiles are available for stabilityai/stable-diffusion-3.5-large with FP8 precision.
Release 1.3.0#
New Supported Models#
microsoft/trellis
Release 1.2.0#
New Supported Models#
black-forest-labs/flux.1-kontext-dev
stabilityai/stable-diffusion-3.5-large
Release 1.1.1#
Summary#
Weight streaming for T5 TRT engine is now available for the black-forest-labs/flux.1-dev model (model profiles for GPUs with 24GB VRAM are updated).
Release 1.1.0#
New Supported Models#
black-forest-labs/flux.1-schnell
New Features#
FP8 and FP4 quantized profiles for black-forest-labs/flux.1-schnell
Memory utilization profiling and automated offloading policy selection based on the host system’s resources
Support for T5 TRT engine with weight streaming for black-forest-labs/flux.1-schnell
Release 1.0.1#
Summary#
New model profiles for black-forest-labs/flux.1-dev
Release 1.0.0#
Summary#
This is the first general availability release of NVIDIA NIM for Generative AI.
New Supported Models#
black-forest-labs/flux.1-dev
New Features#
Dynamic resolution
TRT customization on the user side
FP8 and FP4 quantized profiles for black-forest-labs/flux.1-dev
Support for running on Windows with WSL is beta maturity