Thermostatic/neuraltranslate-27b-mt-nah-es-v1.2
VISIONConcurrency Cost:2Model Size:27BQuant:FP8Ctx Length:32kPublished:Jun 22, 2025License:mitArchitecture:Transformer0.0K Open Weights Cold
Thermostatic/neuraltranslate-27b-mt-nah-es-v1.2 is a 27 billion parameter machine translation model developed by Irving Ernesto. Fine-tuned from Gemma 3 27b Instruct, this model specializes in translating Spanish to Nahuatl. It was trained for 10 epochs on the Axolotl Spanish-Nahuatl Dataset, focusing on this specific language pair.
Loading preview...
Model Overview
Thermostatic/neuraltranslate-27b-mt-nah-es-v1.2 is the first official release of the NeuralTranslate 27b Machine Translation model, developed and funded by Irving Ernesto. This large language model is specifically designed for translation between Spanish and Nahuatl.
Key Capabilities
- Specialized Machine Translation: The model is fine-tuned from Gemma 3 27b Instruct, focusing exclusively on Spanish to Nahuatl translation.
- Targeted Training: It was trained for 10 epochs on the Axolotl Spanish-Nahuatl Dataset, optimizing its performance for this particular language pair.
Limitations and Recommendations
- Language Specificity: This model is strictly for Spanish to Nahuatl translation and will not work for other language pairs (e.g., Nahuatl to English).
- Reliability: The developer notes that even for its intended use, the model's reliability for Spanish to Nahuatl translation is not guaranteed.
- Inference Settings: For optimal performance, it is recommended to use Gemma 3 inference settings:
temperature = 1.0,top_p = 0.95,top_k = 64.