Independent RegistryRTXSPARKS
Community Benchmarks

Submit a benchmark result.

Every submission flows through the RTXsparks validation pipeline: intake → lab review → verified (or rejected, with a reason). Verified results feed into the public SCS / ARS / TLS score updates.

  1. 1. Pending intake
  2. 2. Lab review
  3. 3. Verified

New submission

sub-0142verified
2026-11-08

MSI Spark Titan Desktop

Llama-3.1-70B Q4 throughput — 26.4 tok/s · model llama31-70b

by @vllm-hobbyist

vLLM 0.6, batch=1, 4k ctx

sub-0141reviewing
2026-11-07

Dell Precision S16 RTX Spark

Agent stress (24h) — 0 crashes · model openshell-nemo

by @labpartner

p99 thermal 84°C, no throttle events

sub-0140pending
2026-11-05

ASUS ZenBook 14 Spark

Qwen3-Coder-32B latency — 412 ms first token · model qwen3-coder-32b

by @anonymous

sub-0139rejected
2026-11-04

HP Z-Spark Mini Cluster Node

Sustained TFLOPS — 1340 TFLOPS · model —

by @cluster-ops

Used peak burst window; not sustained.