turningpoint-ai/VisualThinker-R1-Zero

Full fine-tune vs Qwen/Qwen2-VL-2B. Fine-tune: 100% of the weights changed, across 32 of 32 layers.

Tensors changed: Embeddings 2/2, Norms 111/187, MLP 215/216, Attention 324/324

2 derivatives checked

• mradermacher/VisualThinker-R1-Zero-i1-GGUF: Quantization (fewer layers)
• mradermacher/VisualThinker-R1-Zero-GGUF: Same weights, converted (fewer layers)

On the Hub · Downloads on Model Pulse