Hcompany/Holo2-30B-A3B

Full fine-tune vs Qwen/Qwen3-VL-30B-A3B-Thinking. Fine-tune: 100% of the weights changed, across 48 of 48 layers.

Tensors changed: Embeddings 4/4, Norms 266/309, MLP 263/268, Attention 300/300, lm_head 1/1

2 derivatives checked

• mradermacher/Holo2-30B-A3B-GGUF: Quantization
• mradermacher/Holo2-30B-A3B-i1-GGUF: Quantization

On the Hub · Downloads on Model Pulse