Build the fastest configuration that fits.
Map the architecture, expose the communication path, and compare the physics ceiling with benchmark-calibrated reality.
Model execution map
See the active model path, what each device stores, and where devices communicate.
Performance and fit
Physical topology Inspect links and utilization after understanding the logical deployment
Per-device utilization chart Memory, compute, local bandwidth, and network
Published benchmarks for reference click to expand
Real-world throughput plus official model-card task scores from vendor, community, and research sources. Throughput rows are used for configuration alignment; task-score rows are for comparison only.
| Model | Quantization / Mode | Runtime / Benchmark | Hardware / Task | Batch / Eval | Seq / Setting | Batch Rate / Score | Single Rate / Score | Source |
|---|
Browse the ecosystem, not a dropdown.
Base models, fine-tunes, quantizations, active parameters, and Localmaxxing performance signals in one catalog.
Ideal physics, checked against real machines.
The gold set keeps only Localmaxxing runs that map cleanly to a known model, device, quantization, runtime, and command.
Measured versus predicted
Reproducible reference runs
| Model / setup | Observed | Ideal | Calibrated | Error |
|---|
Turn a measured run into headroom.
Use the active plan, enter the speed you measured, and get the implied efficiency, ideal ceiling, and highest-value changes.
Your active plan supplies the model and hardware.
We will compare the entered result with both its physical ceiling and the peer-calibrated expectation.
One payload in, an optimization envelope out.
The same result shape can power a public endpoint once the static site gains a small backend.
{
"observedTokS": 32.2,
"configuration": "active-plan",
"return": ["idealTokS", "expectedTokS", "headroom", "recommendations"]
}