Every model Continuum runs, measured on LM Studio and vLLM.
A page per model: how it performs on LM Studio and on vLLM, what Continuum's durability costs it, the settings it needs, and the quirks it showed under measurement.
UPDATED 2026-09-28 · VERSION 0.1
The models
Every model Continuum runs gets a page of its own, measured the way the benchmarks are: on LM Studio and on vLLM, on one workstation, with every figure recomputed from the run's raw data. Each page carries what was measured on that model: its speed, its quality where it was judged, what a durable workflow adds to its calls, the settings it runs under, and the quirks it showed along the way.
The lineup
Each model is measured on both runtimes, on one machine.
Nemotron 3 Nano 4B
Continuum's inference model on LM Studio and vLLM: quality, latency, throughput under load, and its settings.
Nemotron 3.5 Lightning 30B-A3B
Continuum's extraction model on both runtimes: speed, judged extraction, the settings that extract the most, and its quirks.
No model matches that search, so try fewer words.
House rules
A model is compared only with itself, on the same machine.
Each model runs on both runtimes, each serving the build it runs best, on the same machine and one request at a time unless its page says otherwise, so a difference between the runs comes from the runtime and its build. The machine and the runtime versions are on the benchmarks landing; each model page carries its own settings, its quirks and the commands to take its runs again.