vela3-flash

Vela 3.0 on d3-flash: a 9B decision model with trained adapters for the built-in questions and spans.

Property	Value
Repository	vllm-sr/vela3-flash, pinned at af48828de3f9
Family	Vela 3.0
Parameters	8.83B loaded
Backbone	Qwen3.5 text
Base package	vllm-sr/d3-flash at 8b35c853b90e
Precision (--dtype auto)	BF16 on a GPU, FP32 on a CPU
Weights	about 17.7 GB at BF16, 35.3 GB at FP32
Device memory	32 GiB or more
Reads	text, image, video

Question types

Type	Answers
choice	Yes
noul	Yes
score	Yes
set	Yes, asked as one noul per label
span	Yes, on the router span head

Rules that read the request's images (input.image) or the current turn's videos (input.video) can resolve to it.

Built-in questions

The Vela 3.0 family was trained on the Router's built-in questions, so the Router asks each one in its trained form; the support matrix lists which.

备注

This model has no calibrated thresholds: a built-in question without a predicate uses 0.5, and the configuration load warns with uncalibrated_preset.

Hardware

Each time it loads, the model compares its answers with the reference answers recorded for its revision. They are recorded for:

• AMD GPU (ROCm) (rocm)

On a CPU or an NVIDIA GPU no reference is recorded, so the model serves there with its self-check reported as unverified.

Choose a model lists the validated hardware.

Quantization

The runtime loads no quantized packages of the Vela 3.0 family: it runs at its published precision.

Deploy it

Serve it on its own in Engine mode, to ask it directly:

CPU

vllm-sr serve vllm-sr/vela3-flash --engine --platform cpu

Serve it on a CPU from the CPU image, as --platform cpu does: the ROCm image's PyTorch has no CPU LAPACK, which this backbone needs there.

AMD GPU

vllm-sr serve vllm-sr/vela3-flash --engine --platform rocm

NVIDIA GPU

vllm-sr serve vllm-sr/vela3-flash --engine --platform cuda

Or make it the Router's decision model:

config.yaml

global:
  model_catalog:
    deployments:
      primary:
        provider: model_runtime
        artifact: vllm-sr/vela3-flash
        device: auto
    system:
      decision_model: {deployment: primary}

device: auto takes the first validated GPU with enough free memory, else the CPU; name rocm:0 or cuda:0 to pin a GPU.

To run it on another host and attach the Router to it:

vllm-srun serve vllm-sr/vela3-flash --device auto --host 0.0.0.0 --port 8100

Then set endpoint: http://<host>:8100 and model: vela3-flash on the deployment; see Attach to a runtime you run.

Learn more

• Model card on Hugging Face
• vela3-d3-parity record
• Vela 3.0: the family's sizes and how they differ