Skip to content
Computer vision

Script2Play-4B

a model for playing games from screen frames

1 min read

Script2Play-4B is a vision-language-action (VLA) policy built on Qwen3.5-4B-Base, trained to play video games directly from raw screen capture. It takes a single frame and a text prompt as input, outputting a sequence of keyboard and mouse actions for the next 200 milliseconds. Trained via behavior cloning on Minecraft and Roblox demonstrations, it is well-suited for researchers and developers building interactive agents and autonomous game-playing systems.

Key features

  • Built on the Qwen3.5-4B-Base architecture
  • Trained on 150.3 hours of verified demonstrations (Minecraft and Roblox)
  • Designed to generate low-level mouse and keyboard commands
  • Runs via a standard OpenAI-compatible endpoint in vLLM

What you can do

  • Run the model as a game-playing agent for Minecraft or Roblox via vLLM
  • Integrate the model into custom test environments to evaluate autonomous gameplay
  • Explore vision-language-action model capabilities in real-time control tasks

Example

Launching the model server with vLLM:

vllm serve PATH_TO_THIS_REPO --served-model-name script2play --trust-remote-code \
  --max-model-len 8192 --max-num-batched-tokens 8192 --limit-mm-per-prompt '{"image":1,"video":0}' \
  --attention-backend FLASHINFER --mm-encoder-attn-backend TORCH_SDPA
Read moreDetails
Base model: Qwen/Qwen3.5-4B-Base (approx. 4.54B parameters)
Training: 26.9h of Minecraft and 123.4h of Roblox (no human data or RL used)
Interface: OpenAI-compatible endpoint via vLLM 0.26
Output format: Mouse movement and key presses over a 200 ms window
Discuss Telegram 0

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.

    # GameIntelligenceLab/Script2Play-4B
    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 GameIntelligenceLab/Script2Play-4B --local-dir ~/models/Script2Play-4B
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to Script2Play-4B
  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

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

GameIntelligenceLab/Script2Play-4B
Author
GameIntelligenceLab
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments