typhoon-ai/typhoon-ocr1.5-2b
Full fine-tune vs Qwen/Qwen3-VL-2B-Instruct. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Embeddings 4/4, Norms 211/217, MLP 196/196, Attention 208/208
7 derivatives checked
- mradermacher/typhoon-ocr1.5-2b-i1-GGUF: Quantization (fewer layers)
- mradermacher/typhoon-ocr1.5-2b-GGUF: Same weights, converted (fewer layers)
- nanash66/typhoon-ocr1.5-2b-ROCMFP4-GGUF: Quantization (fewer layers)
- Rattatammanoon/hurricane-ocr-tlpr-v1-LoRA: Adapter
- nanash66/typhoon-ocr1.5-2b-NVFP4: Quantization
- chanasia/typhoon-ocr1.5-2b-GGUF: Quantization (fewer layers)
- mlx-community/typhoon-ocr1.5-2b-8bit: Quantization
2 other repos hold byte-for-byte identical weight files.