Local Qwen3.5-35B-A3B on a 24GB GPU Builds a Full Game From One Spec

Melvin Vivas · X video post · 2026-03-01 · Open on X

Topics: LLM Fundamentals, AI Dev Tools & Productivity, Industry Trends & Job Market · Level: intermediate

Summary

The creator shares a quoted demo in which Qwen3.5-35B-A3B, running locally on a 24GB VRAM GPU, generated a complete game (10 files, 3,483 lines of code) from a single detailed spec prompt. The takeaway is that mid-size open-weight MoE models now handle sizable coding tasks on consumer hardware such as an RTX 3090.

Key points

• Qwen3.5-35B-A3B is a Mixture-of-Experts model (35B total parameters, about 3B active), which is why it runs on a single 24GB consumer GPU.
• Hardware in the demo: 24GB VRAM (the creator plans to try it on an RTX 3090).
• One detailed prompt describing the full game architecture produced 10 files and 3,483 lines of code with no follow-up guidance.
• The spec named enemy types, particle systems, procedural audio, power-ups, boss fights and the player ship.
• Lesson: an upfront, detailed architecture spec lets a local model build a multi-file project in a single pass.

Resources mentioned

• Qwen 3.5 35B · tool · huggingface.co · free
  An open-weight Mixture-of-Experts LLM from Alibaba's Qwen team that can run locally on a 24GB GPU.
  Also in: Running Qwen 3.5 35B Locally in LM Studio for Vision and Prompts (Melvin Vivas on X · notes)
• sudoingX (X account) · person · x.com · free
  The X user whose quoted post shows the local Qwen3.5 game-building experiment.

Try this

• Try running Qwen3.5-35B-A3B locally if you have a GPU with 24GB of VRAM (e.g. an RTX 3090).
• Write one detailed spec of the full app architecture before prompting the model.
• Have a local LLM build a complete browser space-shooter game from one spec prompt, with enemy types, particles, procedural audio, power-ups and boss fights.