--- license: apache-2.0 base_model: - google/gemma-4-E4B-it language: - en pipeline_tag: image-text-to-text library_name: transformers tags: - text-generation-inference - caption-anything - caption - llama-cpp --- # **Gemma4-BLIP3o-Captioner-8B** > **Gemma4-BLIP3o-Captioner-8B** is a fine-tuned image captioning model built on top of **Gemma-4-E4B-it**, designed to mimic and replicate the **BLIP3o (Bootstrapped Language-Image Pretraining) Captioning System** through targeted fine-tuning on ~3K samples from the [BLIP3o-Pretrain-Long-Caption](https://huggingface.co/datasets/BLIP3o/BLIP3o-Pretrain-Long-Caption) dataset. The model features a **modified chat template** with a hardcoded expert system prompt engineered for dense, detail-rich image captioning — covering a wide range of image categories including scenery, natural environments, portraits, objects, and more — with thinking mode disabled by default to prioritize low-latency captioning outputs. It supports **sequential video frame captioning**, making it suitable for temporally ordered visual description tasks. As a captioning-specialized variant, the model may produce artifacts or degraded outputs when used for general-purpose conversational or instruction-following tasks outside its captioning scope. Ideal for applications requiring structured, verbose, and contextually accurate image descriptions in privacy-focused or local inference environments, leveraging the efficient multimodal backbone of Gemma-4-E4B-it. > [!IMPORTANT] > Note: This model is experimental and not an all-purpose one. ## Model Files File Name | Quant Type | File Size | File Link | |-----------|------------|-----------|-----------| | Gemma4-BLIP3o-Captioner-8B.BF16.gguf | BF16 | 15.1 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.BF16.gguf) | | Gemma4-BLIP3o-Captioner-8B.F16.gguf | F16 | 15.1 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.F16.gguf) | | Gemma4-BLIP3o-Captioner-8B.Q2_K.gguf | Q2_K | 4.4 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.Q2_K.gguf) | | Gemma4-BLIP3o-Captioner-8B.Q3_K_L.gguf | Q3_K_L | 5.02 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.Q3_K_L.gguf) | | Gemma4-BLIP3o-Captioner-8B.Q3_K_M.gguf | Q3_K_M | 4.85 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.Q3_K_M.gguf) | | Gemma4-BLIP3o-Captioner-8B.Q3_K_S.gguf | Q3_K_S | 4.65 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.Q3_K_S.gguf) | | Gemma4-BLIP3o-Captioner-8B.Q4_0.gguf | Q4_0 | 5.19 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.Q4_0.gguf) | | Gemma4-BLIP3o-Captioner-8B.Q4_K_M.gguf | Q4_K_M | 5.34 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.Q4_K_M.gguf) | | Gemma4-BLIP3o-Captioner-8B.Q4_K_S.gguf | Q4_K_S | 5.2 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.Q4_K_S.gguf) | | Gemma4-BLIP3o-Captioner-8B.Q5_0.gguf | Q5_0 | 5.69 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.Q5_0.gguf) | | Gemma4-BLIP3o-Captioner-8B.Q5_K_M.gguf | Q5_K_M | 5.76 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.Q5_K_M.gguf) | | Gemma4-BLIP3o-Captioner-8B.Q5_K_S.gguf | Q5_K_S | 5.69 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.Q5_K_S.gguf) | | Gemma4-BLIP3o-Captioner-8B.Q6_K.gguf | Q6_K | 6.22 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.Q6_K.gguf) | | Gemma4-BLIP3o-Captioner-8B.Q8_0.gguf | Q8_0 | 8.01 GB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.Q8_0.gguf) | | Gemma4-BLIP3o-Captioner-8B.mmproj-bf16.gguf | mmproj-bf16 | 992 MB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.mmproj-bf16.gguf) | | Gemma4-BLIP3o-Captioner-8B.mmproj-f16.gguf | mmproj-f16 | 992 MB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.mmproj-f16.gguf) | | Gemma4-BLIP3o-Captioner-8B.mmproj-q8_0.gguf | mmproj-q8_0 | 560 MB | [Download](https://huggingface.co/prithivMLmods/Gemma4-BLIP3o-Captioner-8B/blob/main/Gemma4-BLIP3o-Captioner-8B.mmproj-q8_0.gguf) | ## llama.cpp LLM inference in C/C++ — https://github.com/ggml-org/llama.cpp