d3-lite-GGUF

d3-lite is a 0.85B-parameter multimodal decision model from the vLLM Semantic Router team. It is built on Qwen3.5-0.8B, includes a 0.10B vision encoder, and is released under Apache-2.0. It takes text or JSON input, optionally with several images and videos. It answers multiple questions about that input in a single call: Choice, Yes/No and Score. It returns a probability for every option and generates no text. On the official Jev Decision Index 0.3.2 text board it scores 32.3, first among entries of similar size and on the Pareto frontier, which is 14.9 points above Decision 2.0's 0.8B model. On the vision board it scores 42.1, level with JPT-0.8B, and on the Perception Test video benchmark it scores 59.3. It is also fast. A text request takes a median of 13.6 ms on one AMD Instinct MI325X, and vLLM Semantic Router's Engine mode with batching serves up to 518 requests/s. It can be used through transformers (AutoModel or a pipeline) or served with vllm-sr serve -e vllm-sr/d3-lite.

[!NOTE] This is a Decision Model GGUF that supports the llama.cpp v1/systemone API endpoint.

Model Files

Run with llama.cpp (SystemOne API)

1. Start the Server

Launch llama-server with the GGUF model and wait for the health check endpoint to report ready.

import subprocess, time, requests

MODEL_PATH = "/data/llama.cpp/gguf-new/d3-lite.BF16.gguf"
BASE_URL = "http://127.0.0.1:8080"

server = subprocess.Popen(
    [
        "/data/llama.cpp/build/bin/llama-server",
        "-m", MODEL_PATH,
        "--host", "127.0.0.1",
        "--port", "8080",
        "-ngl", "99",
    ],
    stdout=open("/data/llama-server.log", "w"),
    stderr=subprocess.STDOUT,
)

for _ in range(120):
    try:
        if requests.get(f"{BASE_URL}/health", timeout=2).status_code == 200:
            print("server ready")
            break
    except requests.exceptions.RequestException:
        pass
    time.sleep(2)
else:
    print("server not ready, check /data/llama-server.log")

2. Query the /v1/systemone Endpoint

Send a POST request with a state context and structured questions (choice, noul, or score).

curl -s http://127.0.0.1:8080/v1/systemone \
  -H "Content-Type: application/json" \
  -d '{
    "state": "Customer message: I was charged twice for my order last week and nobody has replied.",
    "questions": {
      "route": {
        "type": "choice",
        "instructions": "Which team should handle this?",
        "criteria": {
          "billing": "payments, charges, refunds, invoices",
          "shipping": "delivery, tracking, lost or late parcels",
          "technical": "bugs, errors, login problems"
        }
      },
      "angry": {
        "type": "noul",
        "instructions": "Is the customer angry?"
      },
      "urgency": {
        "type": "score",
        "instructions": "How urgent is this?",
        "criteria": ["can wait", "this week", "today", "right now"]
      }
    }
  }' | python3 -m json.tool

3. Example Response

{
    "model": "/data/llama.cpp/gguf-new/d3-lite.BF16.gguf",
    "answers": {
        "route": {
            "type": "choice",
            "choice": "billing",
            "probabilities": {
                "billing": 0.9933246348741367,
                "shipping": 0.0009911821477183243,
                "technical": 0.0056841829781450635
            },
            "confidence": 0.9609541327957913
        },
        "angry": {
            "type": "noul",
            "noul": 0.8680201767445472
        },
        "urgency": {
            "type": "score",
            "score": 0.31579992043782523,
            "legend": {
                "0": "can wait",
                "1": "this week",
                "2": "today",
                "3": "right now"
            },
            "probabilities": {
                "0": 0.7724334502730623,
                "1": 0.1750157852323441,
                "2": 0.016868158278299888,
                "3": 0.03568260621629379
            },
            "confidence": 0.5006286367950428
        }
    },
    "usage": {
        "input_tokens": 350,
        "output_tokens": 0
    }
}

llama.cpp

LLM inference in C/C++ — https://github.com/ggml-org/llama.cpp

Quantized using llama.cpp by prithivMLmods.