{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreibq57vb4obhnejdox5bzosw53bfofut3nstsxnzsgy2vocpwgozwq",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mp2kh6i7v6j2"
  },
  "path": "/t/best-model-for-local-usage-and-wokring-on-unity-with-mcp-at-12-gb-vram/177118#post_3",
  "publishedAt": "2026-06-24T17:06:26.000Z",
  "site": "https://discuss.huggingface.co",
  "textContent": "if your running a local LLM alongside a heavy game engine, you should think in terms of small, dense coding models rather than massive MoE models.\n\nThe Best All-Rounder: `Qwen2.5-Coder-7B-Instruct`It is widely considered the gold standard for local, sub-10B coding models. It has excellent comprehension of C# and Unity API patterns, handles logical debugging well, and fits comfortably in your VRAM. I’d go ahead with the Q5 quant if you can get it",
  "title": "Best model for local usage and wokring on Unity with MCP at 12 gb VRAM"
}