ACCELERATE YOUR
AI PERFORMANCE TESTING

New Integration

Now Supporting NVIDIA DGX™ Spark

Metrum Insights now fully supports NVIDIA DGX™ Spark, enabling accelerated AI performance testing on the industry's most advanced AI infrastructure.

v3.5.2 Release

Latest Features Available

Enhanced visualization, multi-node inference, and expanded GPU support

Enhanced Data Visualization with Pulse Support for sharing and publishing data views, role-based access control, dynamic filtering, and UI/UX enhancements

Multi-Node Mesh Inference Workloads Run workloads across multiple nodes with aggregated performance insights and NVIDIA GPU support

Upstream vLLM for AMD GPUs Support for upstream vLLM v0.10.2 with ROCm 7.0 for model serving on AMD GPUs

Fully Automated
AI Performance Benchmarking

Configure unlimited combinations of models, software, hardware, and hyperparameters in seconds.

PARAMETERS

Concurrency Levels
Input and Output Token Lengths
Image Resolutions
Request Rates
Randomized Prompts

CHIPS

NVIDIA Datacenter GPUs
AMD Instinct
Intel Gaudi 3
AMD EPYC
Intel Xeon
NVIDIA RTX GPUs
Intel Arc GPUs
Intel Core Processors
Emerging XPUs

MODELS

GPT-OSS
DeepSeek
Gemma
Phi
Qwen
Mistral
Nemotron
Llama 4+
and more

PIPELINES

RAG
Agents
Inference
Training
Multimodal

MODEL SERVERS

vLLM
TensorRT
SGLang
NIM
Dynamo

HYPERPARAMETERS

Concurrency Levels
Input and Output Token Lengths
Image Resolutions
Request Rates
Randomized Prompts

CHIPS

NVIDIA Datacenter GPUs
AMD Instinct
Intel Gaudi 3
AMD EPYC
Intel Xeon
NVIDIA RTX GPUs
Intel Arc GPUs
Intel Core Processors
Emerging XPUs

MODELS

Llama 4+
DeepSeek
Gemma
Phi
Qwen
Mistral
Nemotron
Falcon
and more

PIPELINES

RAG
Agents
Inference
Training
Multimodal

MODEL SERVERS

vLLM
TensorRT
SGLang
NIM
Dynamo

AI-Powered Analysis with Creator

creator

Smart Hardware Recommendations
with Hardware Sizer

hardware-sizer

Performance Visualization with Pulse

pulse