No official torrent yet β requesting it tells the flock what to host next. You cannot seed a catalog-only page.
Cataloged from vllm-sr/d3-nano
All models by vllm-sr β[Image blocked: d3-nano: Decision 3.0]
d3-nano is the 2B multimodal foundation decision model of Decision 3.0, the decision models of vLLM Semantic Router. Give it an input (text or JSON, optionally with images and videos) and the questions you need answered: pick one of several options, say yes or no, or rate on a scale. It answers them all in one call and returns a probability for every answer, without generating text.
batching profile); the default exact profile gives AutoModel's answers byte for byte when both use the same kernel settings (Serving performance).Images go before the text of the request, each read at up to 1.6 megapixels; videos follow the images, read at 2 frames per second (at most 32 frames spread over the whole video, each at up to 0.2 megapixels). Every question of the request sees all of them.
vLLM Semantic Router serves Decision 3.0 natively in Engine mode, which runs one model behind the System One API on your own machine. Its answers match this package's own d3_server.py exactly, for text, images and videos (measured on AMD Instinct MI325X).
The platform is detected automatically; choose one with --platform auto|cpu|cuda|rocm (default auto).
Engine mode and Decision 3.0 are in the development channel (--channel dev) until a stable vLLM Semantic Router release includes them. The first start pulls the image for your machine and downloads the model. Then ask questions about a request:
Send images as base64 data URLs (PNG, JPEG or WebP) in images and videos (MP4, WebM, QuickTime or Matroska) in videos; every question sees them:
vllm-sr stop stops it. See Router and Engine modes for limits and options.
On one AMD Instinct MI325X, vLLM Semantic Router serves d3-nano at 318 requests/s with p99 under 1 s, 4.3Γ AutoModel (one-question text, batching profile), and its default exact profile gives AutoModel's answers byte for byte when both use the same kernel settings.
[Image blocked: Serving throughput of d3-nano on one AMD Instinct MI325X: AutoModel, the package server and vLLM Semantic Router Engine mode]
AutoModel runs in your own process, one call at a time (the Quickstart code); the package server is its python d3_server.py; Engine mode is vllm-sr serve -e with its default exact profile, or with --runtime-profile batching.
Where Engine mode helps, and what it costs:
exact) answers 16 text questions in 54.2 ms (AutoModel: 57.7 ms), 8 questions about one image in 452 ms (519 ms) and a 32k-token text in 445 ms (477 ms); batching runs the 8 image questions together in 204 ms, 61 % less than AutoModel.exact, about the same as the package server's 16.4 GiB, and 22.7 GiB with batching; AutoModel, serving one request at a time, peaked at 9.1 GiB. Engine mode keeps up to 2 GiB of image features cached.exact) returned answers byte-identical to AutoModel's on all 7 request types (text with 1, 4 or 16 questions, an image with 1 or 8, the video and a 32k-token text) when AutoModel used the same kernel settings: 140 Engine replies and 21 AutoModel calls compared. With the kernel autotuning AutoModel runs in each new process, 1 of the 7 was byte-identical and no decision changed (largest probability difference 0.011). batching changed 0.153 % of decisions on a 4,000-request sample of the public suite (45 of 29,339 questions, near-ties), with gold accuracy 87.707 β 87.744 %.A single short request is fastest in your own process: AutoModel answers one text question 2.8 ms sooner than Engine mode and one new image 26 ms sooner, the cost of going through the router to the model worker and its media path. Engine mode's gains come from serving many requests at once.
One request at a time, warm (p50 / p90, ms):
Requests per second on one GPU, closed loop with up to 64 clients (p99 under 1 s, except *):
β Back-to-back calls in one process (1,000 / mean latency). * p99 over 1 s at every load; the highest rate is shown.
Measured on 2026-10-13, one AMD Instinct MI325X per replica and each side alone on the GPU: ROCm 7.2 with PyTorch 2.12; AutoModel and the package server with Transformers 5.17.0; vLLM Semantic Router Engine mode built from source, with serving optimizations that are newer than its published images. Requests: a support ticket of about 330 tokens with 1, 4 or 16 questions; a 1,456 Γ 1,092 PNG (1.6 MP) with 1 or 8 questions; a 10-second 1080p MP4, read at 2 frames per second; a 32k-token text; every image and video request carries new bytes. One request at a time: 3 warm-up calls, then up to 20 timed calls. Under load: 1, 2, 4, β¦ 64 closed-loop clients, each level warmed up and then counted for 15 to 60 s; Engine mode admits 32 requests in flight per worker and answers the rest at once with 503. Start-up: model files on local disk; Engine mode from the start of its router process, without the CLI's image pull.
On the official text board, d3-nano is on the Pareto frontier, where no entry with as many parameters or fewer scores higher, and ranks first among the 12 entries within 0.75 to 1.33 times its size. LiquidAI d1-3B, a larger model with 3.1 billion parameters, also scores 38.9. Five of the six Decision 3.0 models are on that frontier; d3-edge, second among the entries of its size, is not.
[Image blocked: Jev Decision Index against model size]
[Image blocked: Jev Decision Index by area: d3-nano and Decision 2.0 (2B)]
On the official vision board, the six Decision 3.0 models form the whole Pareto frontier: no other entry is on it. d3-nano is ahead of Intern-Decision-2B, the only other entry within 0.75 to 1.33 times its size.
d3-nano on the public vision benchmarks (β approximate rebuild):
Multiple-choice video question answering (three options per question, chance 33.3): top-1 accuracy (%), videos read with the defaults above.
Jev Decision Index: official leaderboard, text and vision boards 0.3.2, 2026-10-12; the boards evaluated v3.1.0, which has the same weights as this version. Size bands and the Pareto frontier use the parameter counts listed by the boards. d3-nano on the public vision benchmarks and the Perception Test: internal evaluation.
Apache-2.0 (LICENSE [blocked]). Built on Qwen/Qwen3.5-2B (Apache-2.0).
Trained on AMD Instinct MI325X GPUs.
pip install "transformers==5.17.0" torch torchvision pillow opencv-python-headless safetensors acceleratepip install flash-linear-attention # optional: fast GPU kernels for the linear-attention layersimport json
from huggingface_hub import hf_hub_downloadfrom transformers import AutoModel
model = AutoModel.from_pretrained("vllm-sr/d3-nano", trust_remote_code=True)
# Textresult = 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))
# Text and an image (or several)receipt = hf_hub_download("vllm-sr/d3-nano", "assets/example-receipt.png")result = model.system_one( state="The customer says the blender arrived cracked and attached the receipt.", images=[receipt], # local paths, http(s) URLs, PIL images or base64 data URLs 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" } }, "on_receipt": { "type": "noul", "instructions": "Does the receipt list the blender?" }, "payment": { "type": "choice", "instructions": "How was the order paid?", "criteria": { "card": None, "cash": None, "gift card": None } } },)print(json.dumps(result["answers"], indent=2))
# Text and a video (or several)clip = hf_hub_download("vllm-sr/d3-nano", "assets/example-video.mp4")result = model.system_one( state="A short clip from a test camera.", videos=[clip], # local paths, http(s) URLs, base64 data URLs or frame arrays questions={ "direction": { "type": "choice", "instructions": "Which way does the square move?", "criteria": { "right": "From left to right", "left": "From right to left", "still": "It does not move" } }, "color_change": { "type": "noul", "instructions": "Does the square change color?" } },)print(json.dumps(result["answers"], indent=2))
# Or as a pipeline:# transformers.pipeline("decision", model="vllm-sr/d3-nano", trust_remote_code=True)(state=..., questions=..., images=..., videos=...)curl -fsSL https://vllm-sr.ai/install.sh | bash -s -- --channel dev --no-launchexport PATH="$HOME/.local/bin:$PATH"vllm-sr serve -e vllm-sr/d3-nanocurl -s localhost:8899/v1/systemone -H 'content-type: application/json' -d '{ "model": "vllm-sr/d3-nano", "state": "Write a Python function that merges two sorted lists.", "questions": { "kind": {"type": "choice", "instructions": "What kind of work is this?", "criteria": {"code": "Writing or fixing code", "math": "Mathematics", "chat": "Anything else"}}, "reasoning": {"type": "noul", "instructions": "Does answering this need multi-step reasoning?"} }}'curl -s localhost:8899/v1/systemone -H 'content-type: application/json' --data-binary @- <<EOF{ "model": "vllm-sr/d3-nano", "state": "What does the attached image show?", "images": ["data:image/png;base64,$(base64 -w0 chart.png)"], "questions": { "kind": {"type": "choice", "instructions": "What kind of image is this?", "criteria": {"chart": "A chart or plot", "photo": "A photograph", "document": "A document"}} }}EOFcurl -s localhost:8899/v1/systemone -H 'content-type: application/json' --data-binary @- <<EOF{ "model": "vllm-sr/d3-nano", "state": "What happens in the clip?", "videos": ["data:video/mp4;base64,$(base64 -w0 clip.mp4)"], "questions": { "moving": {"type": "noul", "instructions": "Does something move across the scene?"} }}EOF@misc{d3_nano_2026, title = {{d3-nano}: A Multimodal Foundation Decision Model}, author = {{vLLM Semantic Router Team}}, year = {2026}, howpublished = {\url{https://huggingface.co/vllm-sr/d3-nano}}}