jamesdborin/Qwen3-30B-A3B-4layers
This model is derived from Qwen/Qwen3-30B-A3B with:
num_hidden_layers = 4
- Freshly initialized weights (⚠️ not the original 30B weights)
- Total parameters: 3,114,814,464
This checkpoint is suitable for:
- research
- fine-tuning
- architecture experiments