FastToolHub WIRE AI & tool neologism radar
Back home

Intel desk gamescaling

GameScaling

Original link

01

Intro

GameScaling is a recently tracked product, model, or tool in public channels. The radar excerpt reads:

"Tsinghua × miHoYo are open-sourcing GameScaling -- the first open-source generalist game-playing LLM series, with 8 models ranging from 0.8B to 27B parameters." (first-author post 2026-10-22 09:17Z ≈ 17:17 BJT; Chinese recap by @huang18825 7.5k views/99 likes; still circulating through 23:55Z). miHoYo Honkai AI R&D + Tsinghua continued pre-training of Qwen3.5 VLMs on keyboard/mouse gameplay; outputs the next 200 ms of actions; plays Minecraft, Doom, escape rooms; S1 (Minecraft) / S3 (multi-game) checkpoints, Apache-2.0 weights + inference code, no training data; research-only.

The recorded source label is X (search_posts_all) + Hugging Face THUSI-Lab + gamescaling.github.io. Use the public URL below to verify the original post or page.

02

Explain

Based only on the excerpt, GameScaling should be understood from the facts stated there. This write-up does not invent funding amounts, model sizes, or leaderboard scores that are absent from the excerpt or URL metadata.

Confirmable points from the excerpt:

  • "Tsinghua × miHoYo are open-sourcing GameScaling -- the first open-source generalist game-playing LLM series, with 8 models ranging from 0.8B to 27B parameters." (first-author post 2026-10-22 09:17Z ≈ 17:17 BJT; Chinese recap by @huang18825 7.5k views/99 likes; still circulating through 23:55Z).

  • miHoYo Honkai AI R&D + Tsinghua continued pre-training of Qwen3.5 VLMs on keyboard/mouse gameplay; outputs the next 200 ms of actions; plays Minecraft, Doom, escape rooms; S1 (Minecraft) / S3 (multi-game) checkpoints, Apache-2.0 weights + inference code, no training data; research-only.

The radar score for this entry is 0.86 (an internal ranking field, not a third-party benchmark score).

03

Usage

To learn more or try GameScaling, start from public sources:

  1. Open the source link: https://huggingface.co/THUSI-Lab/GameScaling and check the official description and availability.

  2. Follow any access path mentioned in the excerpt (API, open weights, waitlist, in-product entry) using official instructions.

  3. Before production use, verify license terms, regional availability, and safety limits yourself; do not assume quotas, prices, or eval scores that the excerpt does not state.