tasks/seed-quantize-eval.json

{
  "id": "seed-quantize-eval",
  "title": "Quantize a small LLM and measure the quality/latency tradeoff",
  "description": "Take a 1-3B instruct model, produce 4-bit and 8-bit GGUF quantizations, and benchmark tokens/sec on CPU plus accuracy on a small MMLU subset. Summarize the tradeoff in a short report.",
  "artifacts": [
    {
      "type": "model",
      "repo_id": "Qwen/Qwen2.5-1.5B-Instruct"
    }
  ],
  "author": "abidlabs",
  "author_name": "abidlabs",
  "author_avatar": "https://huggingface.co/api/users/abidlabs/avatar",
  "created_at": "2026-06-15T16:28:15+00:00"
}