Your infrastructure.Our benchmarking platform.
You pay for the platform, never the compute. Benchmarks run against cloud instances, on-premises servers, and endpoints you already control, and your infrastructure bill goes straight to your provider.
- Seat
- One account that can create projects and run benchmarks. Read-only members are free.
- Target
- One registered endpoint or hardware configuration available to benchmark against.
Multimodal
Starting from
$30,000/yr
- ASR, VLM, and image generation workloads
- 45 targets included
- Dedicated Solutions Engineer
Pro
Starting from
$90,000/yr
- All Power Tools included
- 75 targets and the Open-Weight Leaderboard
- Dedicated CSM and Solutions Engineer
Enterprise
Starting from
$180,000/yr
- On-premises self-managed install
- Unlimited targets under fair use
- Dedicated TAM and Solutions Engineer
Add-ons
Expand your plan with workload packs. Available from Multimodal tier. Contact sales@metrum.ai to add any of these to your subscription.
Image Generation
SD and FLUX image generation benchmarking. Measures throughput, latency, and quality across diffusion model serving configurations.
Available from Multimodal tier.
LoRA Fine-tuning
LoRA fine-tune benchmarking and validation. Measures adapter training throughput and convergence across hardware configurations.
Available from Multimodal tier.
Versus
Side-by-side model and hardware comparison. Run two configurations in parallel and compare results in a shared reporting view.
Available from Multimodal tier. Included in Pro and above.