Vela-2.0-0.8B

The smallest Vela 2.0 decoder. On a GPU it is ahead of the Vela 1.0 specialists on most built-in signals; on a CPU a request takes seconds.

Property	Value
Repository	vllm-sr/Vela-2.0-0.8B, pinned at a778eb2ae230
Family	Vela 2.0
Parameters	0.756B loaded
Backbone	Qwen3.5 text
Precision (--dtype auto)	BF16 on a GPU, FP32 on a CPU
Weights	about 1.5 GB at BF16, 3.0 GB at FP32
Device memory	4 GiB or more
Reads	text

Question types

Type	Answers
choice	Yes
noul	Yes
score	Yes
set	Yes
span	Yes, on the router, broad span heads

Built-in questions

The Vela 2.0 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.

A rule without a predicate takes this model's calibrated threshold:

Preset	Threshold
jailbreak (jailbreak)	0.71
domain (domain)	0.38
pii (pii)	0.07
fact check (fact_check)	0.994
user feedback (feedback)	0.34

Hardware

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

• CPU (cpu)
• NVIDIA Hopper (sm_90) (cuda:sm_90)
• AMD GPU (ROCm) (rocm)

Where a hardware reference and its device class's are both recorded, the hardware one is checked.

Choose a model lists the validated hardware.

Quantization

The runtime loads no quantized packages of the Vela 2.0 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/Vela-2.0-0.8B --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/Vela-2.0-0.8B --engine --platform rocm

NVIDIA GPU

vllm-sr serve vllm-sr/Vela-2.0-0.8B --engine --platform cuda

Or make it the Router's decision model:

config.yaml

global:
  model_catalog:
    deployments:
      primary:
        provider: model_runtime
        artifact: vllm-sr/Vela-2.0-0.8B
        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/Vela-2.0-0.8B --device auto --host 0.0.0.0 --port 8100

Then set endpoint: http://<host>:8100 and model: Vela-2.0-0.8B on the deployment; see Attach to a runtime you run.

Learn more

• Model card on Hugging Face
• vela2-decision-model-sizes record
• vela2-parity record
• cuda-vela2-performance record
• Vela 2.0: the family's sizes and how they differ