GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
RBDash-Team
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
RBDash-Team
Model Tree
Qwen/Qwen3.8-27B
Input Modalities
Output Modalities
Supported Functionality
| Base model | Qwen/Qwen3.8-27B |
| Architecture | Qwen3.5 vision-language backbone |
| Output head | 255-way linear readout (readout.safetensors) replacing the LM head |
| Attention | Full-attention layers are non-causal; the linear-attention recurrence stays causal |
| Pooling | Last token |
| Calibration | One fitted temperature, 0.8821, stored in decision_config.json |
| Precision | bfloat16 |
| License | Apache-2.0 |
The readout is initialised from the LM-head rows of 255 single-token answer codes (A to JT). The model reads the prompt once, takes the hidden state of the last token, and applies the readout:
text
probs = softmax(readout(hidden[-1])[:n_options] / T)Decision Index edition 0.3, public vision suite (11 benchmarks, 12,210 rows, one to four images per row): public vision index 73.50, with every row answered and no errors. Images were used at native resolution, capped at 1,638,400 pixels each.
This is the public half of the Vision board's score only. The board's Full score also includes private tests that are run by the maintainers, so 73.50 is not a leaderboard rank.
bash
pip install -r requirements.txtpython
from rbdash_decision import RBDashDecision model = RBDashDecision.from_pretrained("RBDash-Team/RBDash-Decision-v0.1") # Multiple choice. List items become options A, B, C, ...r = model.decide("How many organs are in the image?", ["3", "2", "1", "0"], images=["scan.jpg"])print(r["choice"], r["probabilities"]) # Yes/nor = model.decide("Is the meme hateful?", kind="noul", images=["meme.png"])print(r["noul"]) # probability of "yes"Command line:
bash
python rbdash_decision.py --model . --question "How many organs?" \ --option 3 --option 2 --option 1 --option 0 --image scan.jpgrbdash_decision.py depends only on torch, torchvision, transformers, safetensors, pillow and huggingface_hub. requirements.txt also installs flash-linear-attention (pure Python), which speeds up the linear-attention layers; the first run is slow while triton compiles its kernels. A GPU with about 60 GB free memory is needed for bfloat16.
causal-conv1d is optional and kept out of requirements.txt on purpose, because it can need a compiler. After the main install, run pip install --no-build-isolation -r requirements-optional.txt. It downloads a prebuilt wheel from GitHub Releases when one matches your CUDA, torch and Python versions (the v1.7.0 release covers torch 2.6 to 2.10); otherwise it compiles from source, which needs nvcc and takes several minutes. Without it the model still runs through a slower PyTorch path.
The Decision Index kit does not ship an engine for this model, so rbdash_decision_engine.py in this repository plugs it in. The prompt is part of the trained model: keep rbdash_decision.py unchanged, because other wording can change individual predictions.
bash
# 1. Get the model and the kithf download RBDash-Team/RBDash-Decision-v0.1 --local-dir RBDash-Decision-v0.1git clone https://github.com/apolinario/decision-index && cd decision-indexpip install -e '.[vision]'pip install -r ../RBDash-Decision-v0.1/requirements.txt # 2. Build the public vision suite (about 5.7 GB of downloads; Winoground is gated, accept its terms first)hf auth logindecision-index suite rebuild --vision --work workdecision-index vision verify --dir work/vision # 3. Run and score it (one GPU with about 60 GB free)PYTHONPATH=../RBDash-Decision-v0.1 decision-index vision pipeline --dir work/vision \ --engine rbdash_decision_engine:RBDashEngine \ --option checkpoint=../RBDash-Decision-v0.1 \ --out runs/vision-rbdash-decision-v0.1The run checkpoints to runs/vision-rbdash-decision-v0.1/results.jsonl and resumes if interrupted; add --limit 50 for a quick trial. When it finishes, vision-scores.json in the same directory reports the public vision index, and complete must be true. A --limit run scores only the rows it ran, so it reports complete: false and an index near zero; that is expected.
The engine handles one request at a time on one GPU (the kit's default of 8 worker threads is serialised inside the engine). flash-linear-attention comes with requirements.txt and causal-conv1d with requirements-optional.txt; without them the model runs through slower PyTorch fallbacks. Triton compiles its kernels on first use, so the first minutes of a run with flash-linear-attention are slow.
from rbdash_decision import RBDashDecision model = RBDashDecision.from_pretrained("RBDash-Team/RBDash-Decision-v0.1") # Multiple choice. List items become options A, B, C, ...r = model.decide("How many organs are in the image?", ["3", "2", "1", "0"], images=["scan.jpg"])print(r["choice"], r["probabilities"]) # Yes/nor = model.decide("Is the meme hateful?", kind="noul", images=["meme.png"])print(r["noul"]) # probability of "yes"python rbdash_decision.py --model . --question "How many organs?" \ --option 3 --option 2 --option 1 --option 0 --image scan.jpg# 1. Get the model and the kithf download RBDash-Team/RBDash-Decision-v0.1 --local-dir RBDash-Decision-v0.1git clone https://github.com/apolinario/decision-index && cd decision-indexpip install -e '.[vision]'pip install -r ../RBDash-Decision-v0.1/requirements.txt # 2. Build the public vision suite (about 5.7 GB of downloads; Winoground is gated, accept its terms first)hf auth logindecision-index suite rebuild --vision --work workdecision-index vision verify --dir work/vision # 3. Run and score it (one GPU with about 60 GB free)PYTHONPATH=../RBDash-Decision-v0.1 decision-index vision pipeline --dir work/vision \ --engine rbdash_decision_engine:RBDashEngine \ --option checkpoint=../RBDash-Decision-v0.1 \ --out runs/vision-rbdash-decision-v0.1