QORIN Model Weights & Tokenizers

Official model repository for QORIN (The Quiet Origin of Reasoning and Intelligence Networks).

Available Checkpoints

1. qorin-100m (SFT v3)

  • Path: qorin-100m/pytorch_model.pt
  • File Size: 617,289,583 bytes
  • SHA-256: d771ea9f5465e3bcf2b619a32f7797742e500a01704a4d5f582eb5db4adc7510
  • Context Length: 4,096 tokens
  • Architecture: 14 layers, 12 attention heads, 4 KV heads (3:1 GQA), SwiGLU FFN (dim 2048), RoPE theta 10k (scale 8.0).
  • Tokenizer: qorin-100m/tokenizer/ (16,000 BPE vocabulary)

2. qorin-500m (SFT Exp-F)

  • Path: qorin-500m/pytorch_model.pt
  • File Size: 2,078,616,611 bytes
  • SHA-256: b5b7091e951e49263e7b55e78b1f6837b5713ffb2f197c20366f39f3cd9b1e4
  • Context Length: 2,048 tokens
  • Architecture: 20 layers, 24 attention heads, 4 KV heads (6:1 GQA), SwiGLU FFN (dim 3840), RoPE theta 10k (scale 1.0).
  • Tokenizer: qorin-500m/tokenizer/ (32,768 BPE vocabulary)
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Space using vynor/qorin-models 1