Supported models

These are the decision models the built-in model runtime serves: every model the Router can ask a System One question. Each one has its own page, generated from the catalog the Router and the runtime use, so its question types, sizes and pinned revision always match the code.

Which one?

Start with the default, vela3-fast, on the CPU. Move to a larger Vela 3.0 size for more accuracy with the built-in questions still trained, or to Decision 3.0 for your own questions about images and video. Choose a model compares them.

Families

Family	What it is for	Spans	Runs on
Decision 3.0	Your own questions about text, images and video, zero-shot, from 0.6B to 27B	No	GPU
Decision 2.0	Your own questions about text, zero-shot, from 0.6B to 27B	No	CPU (Kai-0.6B) or GPU
Decision 1.0	The first decision models, encoders and decoders from 0.6B to 9B	No	CPU (0.6B) or GPU
Vela 3.0	The Router's own models: trained for the built-in questions, with spans; the default	Yes	CPU (vela3-fast) or GPU
Vela 2.0	The previous Router models, with set and span questions	Yes	CPU (0.3B) or GPU

Every model answers choice and noul questions; the tables below show the rest. A model without spans can still be the decision model: bind pii and hallucination to a span-capable deployment (Span capability).

All models

Decision 3.0

Model	Parameters	Answers	Spans	Reads	Device memory
d3-edge	0.595B	choice, noul, score	—	text, image, video	4 GiB
d3-lite	0.853B	choice, noul, score	—	text, image, video	4 GiB
d3-nano	2.21B	choice, noul, score	—	text, image, video	8 GiB
d3-mini	4.54B	choice, noul, score	—	text, image, video	16 GiB
d3-flash	8.39B	choice, noul, score	—	text, image, video	32 GiB
d3	26.1B	choice, noul, score	—	text, image, video	72 GiB
d3-turbo	26.1B	choice, noul, score	—	text, image, video	72 GiB
d3-ultra	26.1B	choice, noul, score	—	text, image, video	72 GiB

Decision 2.0

Model	Parameters	Answers	Spans	Reads	Device memory
Decision-2.0-Kai-0.6B	0.597B	choice, noul, score	—	text	4 GiB
Decision-2.0-Eos-0.8B	0.753B	choice, noul, score	—	text	4 GiB
Decision-2.0-Sol-2B	1.88B	choice, noul, score	—	text	8 GiB
Decision-2.0-Nox-4B	4.21B	choice, noul, score	—	text	16 GiB
Decision-2.0-Lux-9B	7.94B	choice, noul, score	—	text	24 GiB
Decision-2.0-Vega-27B	29.4B	choice, noul, score	—	text	72 GiB

Decision 1.0

Model	Parameters	Answers	Spans	Reads	Device memory
Decision-1.0-Kai-0.6B	0.572B	choice, noul, score	—	text	4 GiB
Decision-1.0-Lex-0.6B	0.572B	choice, noul, score	—	text	4 GiB
Decision-1.0-Route-0.6B	0.572B	choice, noul, score	—	text	4 GiB
Decision-1.0-Eos-0.8B	0.753B	choice, noul, score	—	text	4 GiB
Decision-1.0-Sol-2B	1.88B	choice, noul, score	—	text	8 GiB
Decision-1.0-Nox-4B	4.21B	choice, noul, score	—	text	16 GiB
Decision-1.0-Lux-9B	7.94B	choice, noul, score	—	text	24 GiB

Vela 3.0

Model	Parameters	Answers	Spans	Reads	Device memory
vela3-fast	1.52B	choice, noul, span	router	text	12 GiB
vela3-lite	0.903B	choice, noul, score, span	router	text, image, video	4 GiB
vela3-mini	4.79B	choice, noul, score, span	router	text, image, video	16 GiB
vela3-flash	8.83B	choice, noul, score, span	router	text, image, video	32 GiB
vela3-turbo	27.6B	choice, noul, score, span	router	text, image, video	72 GiB

Vela 2.0

Model	Parameters	Answers	Spans	Reads	Device memory
Vela-2.0-0.3B	0.309B	choice, noul, score, set, span	router	text	4 GiB
Vela-2.0-0.8B	0.756B	choice, noul, score, set, span	router, broad	text	4 GiB
Vela-2.0-4B	4.21B	choice, noul, score, set, span	router, broad	text	16 GiB
Vela-2.0-9B	7.95B	choice, noul, score, set, span	router, broad	text	24 GiB

"Device memory" is the catalog's minimum for the model's device; measure your own peak with your input lengths and concurrency.

Embedding models

The embedding models that recall features use, Vela 1.0 Embedding, Qwen3-Embedding-0.6B and Vela 1.0 Omni, are not System One models and are not listed here. See Embeddings and Images and audio.

How these pages are made

The model pages are generated by tools/docs/generate_model_pages.py from the Router's decision catalog, the runtime's registry tables, the recorded reference answers and the Router's calibrated thresholds. Run make docs-model-pages after changing a built-in model, and make docs-generated-check fails until you do. The family pages and this page are written by hand around their generated tables.