MoE with 3.8B active parameters
According to the model card, the Mixture-of-Experts totals 25.2B parameters with 3.8B active, using 8 active of 128 experts plus 1 shared expert across 30 layers.
Open Source Model Profile · unsloth
gemma-4-26B-A4B-it
gemma-4-26B-A4B-it is a 26.54B-parameter gemma4 MoE fine-tune from unsloth. According to the model card, it activates 3.8B parameters with 256K text-image context.
- Publisher
- unsloth
- Task
- image-text-to-text
- Model type
- gemma4
- License
- apache-2.0
- Library
- Unknown
- Publication status
- Accepted · not indexed
Model overview
gemma-4-26B-A4B-it is published by unsloth as a gemma4 image-text-to-text fine-tune. The captured configuration identifies Gemma4ForConditionalGeneration and Safetensors metadata reports 26,544,131,376 parameters. According to the model card, it follows google/gemma-4-26B-A4B-it as a 25.2B-total MoE with 3.8B active parameters and 256K context, with card data recording apache-2.0.
Recorded capabilities
256K multimodal input
According to the model card, the model supports 256K-token context with interleaved text-image input and a configurable visual token budget from 70 to 1,120 tokens.
Function calling and system roles
According to the model card, the model documents native function-calling support, native system-role support, and standardized sampling of temperature 1.0, top_p 0.95, and top_k 64.
Unsloth fine-tune record
Hub tags identify google/gemma-4-26B-A4B-it as base and finetune source, and the card documents processor-based text, image, and video loading patterns.
Use cases in the source record
- Interleaved text-image generation using the documented multimodal input pattern and configurable visual token budgets.
- Function-calling and system-prompted conversational workflows using the publisher-documented sampling and thinking-mode controls.
- Multilingual text-image experiments across the documented 35-plus supported languages with processor-based loading.
Limitations and unknowns
- No evaluation results were extracted from this record.
- Provider state is historical snapshot data and should be refreshed before being presented as current.
- No pricing, VRAM, or current availability figures were extracted.
Source and provenance
Source: unsloth/gemma-4-26B-A4B-it
Captured: Unknown. Processed: 2026-09-07T19:36:03.456860+00:00.
Read our How to Run Gemma 4 Guide! See Unsloth Dynamic 2.0 GGUFs for our quantization benchmarks. Gemma 4 can now be run and fine-tuned in Unsloth Studio . Read our guide . See all versions of Gemma 4 (GGUF, 16-bit etc.) in our collection . Hugging Face | GitHub | Launch Blog | Documentation License : Apache 2.0 | Authors : Google DeepMind Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on small models) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants. Gemma 4 features…
F001F002F003F004F005F006F007F008F009F016F017F019F020F021F022F025F027F030