Skip to content
Other

Starlord-4B-v2.0-preview

Qwen-Based Decision Model

2 min read

The new Starlord-4B-v2.0-preview model is built on Qwen/Qwen3.5-4B (4.54 billion parameters) and functions as a decision model. Instead of generating text, it reads candidate logits from the original vocabulary and processes the state along with all questions in a single forward pass. The model supports a context length of up to 131,072 tokens and up to 512 choices per question. This development will benefit researchers and engineers working on logical option selection and benchmarks.

Key features

  • Works without text generation by reading logits from the original vocabulary
  • Supports context up to 131,072 tokens for joint queries
  • Capable of handling up to 512 choices per single question
  • Requires a compatible BF16 CUDA GPU (such as an NVIDIA H800) and FlashAttention

What you can do

  • Run performance evaluations on the JevBench and Decision Index benchmarks
  • Deploy the model as a local service for evaluating response options
  • Use it in research pipelines for multiple-choice tasks

Example

export PYTHONPATH="$PWD/inference${PYTHONPATH:+:$PYTHONPATH}" 
CUDA_VISIBLE_DEVICES=0 python -m decision_index pipeline \
  --edition 0.3 \
  --suite-dir /path/to/suite-0.3 \
  --engine di_engine:DecisionIndexEngine \
  --model . \
  --option maximum_tokens=131072 \
  --out decision-index-results
Read moreDetails
Benchmarks: JevBench Public — 94.37% (218/231), Decision Index Public 0.3 — 78.91%
Hardware: 1 × NVIDIA H800 with BF16 and CUDA Graph support
Limitations: KV-cache disabled, token generation set to 0
Stack: PyTorch 2.10.0, Transformers 5.3.0, FlashAttention 2.8.3
Discuss Telegram 0 24

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 8.5 GB · ready files9.8 GB
  • FP8 / INT8recommendedweights 4.2 GB5.2 GB
  • 4-bitweights 2.4 GB3.1 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 3080

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # xsdfasfgsa/Starlord-4B-v2.0-preview
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download xsdfasfgsa/Starlord-4B-v2.0-preview --local-dir ~/models/Starlord-4B-v2.0-preview
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

  1. Qwen3.5-4B-Basefamily root
  2. Qwen3.5-4B
  3. Starlord-4B-v2.0-preview

New posts about versions of Qwen3.5-4B-Base (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 12All on Hugging Face
Merges 1All on Hugging Face

About the model

xsdfasfgsa/Starlord-4B-v2.0-preview
Author
xsdfasfgsa
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

240

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments