Momix-44/gemma-4-26B-a4b-it-Adamus-SFT

VISIONConcurrency Cost:2Model Size:26BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:Apr 28, 2026Architecture:Transformer Cold

Momix-44/gemma-4-26B-a4b-it-Adamus-SFT is a 26 billion parameter Gemma-4 based instruction-tuned language model, fine-tuned by Momix-44 using Unsloth. This model is available in GGUF format, including a BF16 multimodal version, and is designed for general language tasks. It supports a 32768 token context length, making it suitable for applications requiring extensive context processing.

Loading preview...

Overview

Momix-44/gemma-4-26B-a4b-it-Adamus-SFT is a 26 billion parameter instruction-tuned model based on the Gemma-4 architecture. It was fine-tuned by Momix-44 using the Unsloth framework, which facilitated a 2x faster training process. The model is provided in GGUF format, offering various quantization options including a BF16 multimodal version.

Key Capabilities

  • Instruction Following: Optimized for understanding and executing instructions.
  • Multimodal Support: Includes a BF16-mmproj.gguf file, indicating potential for multimodal applications, though specific capabilities are not detailed.
  • GGUF Format: Ready for use with llama-cli and ollama (with specific conversion steps for vision models).

Good For

  • General Language Tasks: Suitable for a wide range of instruction-based natural language processing applications.
  • Developers using Ollama: Provides clear instructions for integrating the multimodal version into Ollama, despite its current limitations with separate mmproj files.
  • Efficient Deployment: The GGUF format allows for flexible deployment across various hardware configurations.

Limitations

  • Ollama currently requires a specific conversion process for vision models to unify the mmproj files, as it does not natively support separate mmproj files.