Workflow Platform LensLive PulseLive Workspace RelayLive HearthLive FusionLive Tools Docs Studies Contact

WorkspaceTry Lens

XARIV

XARIV helps engineering teams make better AI infrastructure decisions — planning, benchmarking, explainability, optimization, and forecasting in one platform.

Product

Home
Workflow
Platform
Lens
Pulse
Relay
Hearth
Fusion
Tools
Docs
Architecture Studies
Blog
Open Source
Company
Contact

Connect

GitHub
LinkedIn
hello@xarivlabs.com

© 2026 XARIV. All rights reserved.

Optional · Step 0 · Quick check

GPU Memory Calculator

Break down weights, KV cache, activations, and total HBM required.

Model

GPU

Precision

Context length

Batch size

Tensor parallel

LoRA adapters (GB)

Results

Weights: 32.0 GB
KV cache: 21.47 GB
Activations (est.): 1.34 GB
Total (+15% overhead): 63.0 GB
GPUs needed: 1

Optional step 0 — continue to Define workload (Lens) for the full platform workflow.