LLM EXPLORER 68,192 MODELS INDEXED

Gemma4 26B Atc Finetune 32 64 V2 Merged 16bit by xl-24

By xl-24 · 44 downloads

Gemma4 26B Atc Finetune 32 64 V2 Merged 16bit is an open-source language model by xl-24. Features: 26b LLM, VRAM: 51.6GB, License: apache-2.0, Quantized, Fine-Tuned, Merged, LLM Explorer Score: 0.44, ELO: 1438.

  Merged Model   16bit Base model:finetune:unsloth/ge... Base model:unsloth/gemma-4-26b...   Conversational   En   Endpoints compatible   Finetuned   Gemma4   Image-text-to-text   Quantized   Region:us   Safetensors   Sharded   Tensorflow   Unsloth

Gemma4 26B Atc Finetune 32 64 V2 Merged 16bit Benchmarks

nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

Gemma4 26B Atc Finetune 32 64 V2 Merged 16bit Parameters and Internals

LLM NameGemma4 26B Atc Finetune 32 64 V2 Merged 16bit
Repository πŸ€—https://huggingface.co/xl-24/gemma4-26b-atc-finetune-32-64-v2-merged-16bit 
Base Model(s)  Gemma 4 26B A4B It   unsloth/gemma-4-26B-A4B-it
Merged ModelYes
Model Size26b
Required VRAM51.6 GB
Updated2026-11-21
Maintainerxl-24
Model Typegemma4
Model Files  49.9 GB: 1-of-2   1.7 GB: 2-of-2
Supported Languagesen
Quantization Type16bit
Model ArchitectureGemma4ForConditionalGeneration
Licenseapache-2.0
Model Max Length262144
Transformers Version5.17.0
Tokenizer ClassGemmaTokenizer
Padding Token<pad>
Torch Data Typebfloat16

Best Alternatives to Gemma4 26B Atc Finetune 32 64 V2 Merged 16bit

Best Alternatives
Context / RAM
Downloads
Likes
OTel 2.0 Decision 26B A4B 4bit0K / 17.3 GB9147554
...6B A4B It Qat Q4 0 Unquantized0K / 51.6 GB3369650
Gemma 4 26B A4b It 4bit0K / 15.4 GB2754379
...yberSecurity 26B A4B V2 Q8 MLX0K / 40.6 GB3931
...yberSecurity 26B A4B V2 Q4 MLX0K / 35.4 GB2461
...6B A4B It Qat Q4 0 Unquantized0K / 51.6 GB34873
Gemma 4 26B A4b It 8bit0K / 27.9 GB291714
OTel 2.0 Decision 26B A4B 4bit0K / 17.3 GB631
Gemma 4 26B A4b It UD MLX 4bit0K / 16.2 GB152734
Gemma 4 26B A4b It MLX 8bit0K / 27.9 GB93616
Note: green Score (e.g. "73.2") means that the model is better than xl-24/gemma4-26b-atc-finetune-32-64-v2-merged-16bit.