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

KV Cache Calculator

Estimate KV cache memory from model, context, batch, and precision.

Related study →

Model

GPU

Precision

Context length

Batch size

GPUs (replica)

Results

KV bytes / token: 327680 B
KV cache total: 21.47 GB
Per GPU (even split): 21.47 GB
GPU capacity: 80 GB

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