Skip to content
Other

Qwen3.5-9B Rubik's Cube

сборка кубика через tool calls

2 min readupdated

Файнтюн Qwen3.5-9B от arjanssuri, который решает кубик Рубика из реального скрамбла end-to-end. Модель читает кубик как текст, пишет рассуждение в ответе (thinking off) и крутит грани через инструмент move; встроенного солвера, поиска и исполнения кода нет. Размер — 9,65 млрд параметров.

Ключевые особенности

  • Следует CFOP: cross, F2L, two-look OLL/PLL.
  • Каждый шаг — lookup, а не поиск: для F2L одна таблица из 59 алгоритмов вместо 236.
  • Один ход за вызов инструмента; полная сборка — медиана 85 ходов за 22 tool calls.
  • На seed-1: 42/64 полных сборок (66%), 28/32 скрамблов решены хотя бы с одной из двух попыток (88%).
  • Ограничения: одна ориентация (белый низ, зелёный фронт) и один формат промпта/кубика.

Пример

pip install vllm openai
vllm serve arjanssuri/qwen3.5-9b-rubiks-cube --max-model-len 32768 --enable-auto-tool-choice --tool-call-parser qwen3_coder
python play.py --base-url http://localhost:8000/v1 --model arjanssuri/qwen3.5-9b-rubiks-cube
ПодробнееБенчмарки: 42/64 (66%) полных сборок; 88% скрамблов решены в 1 из 2 попыток; F2L — 97%, last layer — 94%.
VRAM не указан; для запуска нужны vLLM, max-model-len 32768, thinking off и temperature 1.0.
Обучение: LoRA SFT на трассах от движка + GRPO-style RL на cross; стадии v3–F5 слиты в базовые веса.
Слабые места: ~1 из 5 six-move crosses и H/Z perms примерно в половине случаев. Полная сборка — около 14K токенов.
Discuss Telegram 0 25

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 18 GB · ready files20 GB
  • FP8 / INT8recommendedweights 9.0 GB10 GB
  • 4-bitweights 5.1 GB6.1 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 2080 Ti · BF16 — from RTX 4090

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.

    # arjanssuri/qwen3.5-9b-rubiks-cube
    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 arjanssuri/qwen3.5-9b-rubiks-cube --local-dir ~/models/qwen3.5-9b-rubiks-cube
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to qwen3.5-9b-rubiks-cube
  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. Qwen3.5-9B-Basefamily root
  2. Qwen3.5-9B
  3. qwen3.5-9b-rubiks-cube

New posts about versions of Qwen3.5-9B-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 6All on Hugging Face
Merges 1All on Hugging Face

About the model

arjanssuri/qwen3.5-9b-rubiks-cube
Author
arjanssuri
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

250

Ask AI about this model

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

Comments