Model reference · open weights

Decision-2.0-Lux

Embeddings vllm-sr Embeddings 1 build Open weights 574 dl/mo

Decision-2.0-Lux is an open-weight embedding model from vllm-sr. Decision-2.0-Lux-9B (BF16) weighs 17 MB; the smallest configuration that runs it is RTX 3060 12 GB.

For self-hosting, compare this build's requirements below with the GPU you plan to rent.

  • Decision-2.0-Lux-9B is a 7.94B parameter feature extraction model by vllm-sr designed to answer choice, yes/no, and scoring questions in a single forward pass.
  • It returns probabilities for every answer without generating text and supports a context length of 16,384 tokens.
  • The model is released under the Apache-2.0 license.

Summary of the vllm-sr/Decision-2.0-Lux-9B model card, 2026-10-08

What it is

Released byvllm-sr
Released2026-09-29
VRAM17 MB for the weights

What it runs on

Memory and cards for Decision-2.0-Lux-9B (BF16)

17 MBweights, file size
1.1 GBruntime overhead
CardRunsMemory
RTX 3060 12 GBfits11.6 GB
RTX 4060 Ti 16 GBfits15.4 GB
RTX 3090 24 GBfits23.4 GB
RTX 4090 24 GBfits23.4 GB
RTX 5090 32 GBfits31.0 GB
L40S 48 GBfits44.0 GB
A100 80 GBfits78.2 GB
H100 80 GBfits78.1 GB
RTX PRO 6000 Blackwell 96 GBfits93.8 GB
DGX Spark (GB10) 128 GB unifiedfits107 GB
H200 141 GBfits138 GB
B200 180 GBfits176 GB

From the model card

What vllm-sr says about Decision-2.0-Lux

Read the model card

Decision-2.0-Lux-9B is the 9B model of Decision 2.0, the decision models of vLLM Semantic Router. Give it an input (text or JSON) and the questions you need answered: pick one of several options, say yes or no, or rate on a scale. It answers them all at once and returns a probability for every answer, without generating text.

Parameters7.94B
Context length16,384 tokens
Decision typesChoice · Yes / No · Score
LicenseApache-2.0

Highlights

  • Top JevArena score of its size: 68.1, ahead of the 2 other same-size models compared.
  • Ahead of Decision 1.0 Lux: +2.3 on JevArena and +2.8 on the Jev Decision Index.
  • Speed: a median of 18.4 ms per single-question request on a single GPU.
  • Many questions, one pass: Choice, Yes / No and Score questions about the same input are answered together in one forward pass, with a probability for every option.

Quickstart

pip install "transformers>=5.17" torch safetensors
import json

from transformers import AutoModel

model = AutoModel.from_pretrained("vllm-sr/Decision-2.0-Lux-9B", trust_remote_code=True)
result = model.system_one(
    state="The order arrived damaged yesterday. The customer has a receipt and asks for a replacement today.",
    questions={
        "route": {
            "type": "choice",
            "instructions": "Which team should handle this request?",
            "criteria": {
                "returns": "Refunds, replacements and damaged deliveries",
                "billing": "Payments, invoices and charges",
                "technical": "Product setup and faults"
            }
        },
        "receipt": {
            "type": "noul",
            "instructions": "Does the customer have a receipt?"
        },
        "urgency": {
            "type": "score",
            "instructions": "How urgent is this request?",
            "criteria": [
                "Routine",
                "Soon",
                "Today"
            ]
        }
    },
)
print(json.dumps(result["answers"], indent=2))

# Or as a pipeline:
# transformers.pipeline("decision", model="vllm-sr/Decision-2.0-Lux-9B", trust_remote_code=True)(state=..., questions=...)

Evaluation

ModelJevArena ↑Human-labelled transfer ↑Jev Decision Index ↑
Decision-2.0-Lux-9B68.156.246.3
Decision 1.0 Lux65.855.843.5
Nimble v262.153.6—

JevArena

Jev Decision Index

License

Apache-2.0 (LICENSE).

Citation

@misc{decision_2_0_lux_9b_2026,
  title        = {{Decision-2.0-Lux-9B}: A Decision 2.0 Model for Structured Decisions},
  author       = {{vLLM Semantic Router Team}},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/vllm-sr/Decision-2.0-Lux-9B}}
}

Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.

© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms