oxalis-ink/shamrock-0-e2b

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:5.1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 20, 2026License:cc-by-nc-4.0Architecture:Transformer Open Weights Featherless Exclusive Cold

shamrock-0-e2b is a 5.1 billion parameter language model developed by Oxalis, fine-tuned from Google DeepMind's Gemma 4 E2B. This model is specifically optimized for language learning applications within the Oxalis desktop app, focusing on Japanese, Korean, and Simplified Chinese. It supports functionalities like camera translate, text translate, tutor, talk, and dictionary, and includes a multi-token-prediction drafter for speculative decoding.

Loading preview...

Overview

shamrock-0-e2b is a 5.1 billion parameter language model developed by Oxalis, fine-tuned from Google DeepMind's Gemma 4 E2B. It is specifically designed for integration into the Oxalis desktop app to enhance language-learning experiences across Japanese, Korean, and Simplified Chinese. The model includes a multi-token-prediction (MTP) drafter for speculative decoding, available in GGUF and Hugging Face formats.

Key Capabilities

  • Specialized Language Learning: Optimized for core language-learning surfaces such as camera translation, text translation, tutoring, conversational practice, and dictionary lookups.
  • Multilingual Focus: Tailored for high performance in Japanese, Korean, and Simplified Chinese.
  • Efficient Inference: Includes an MTP drafter for speculative decoding, enhancing inference speed.
  • GGUF Compatibility: Provided with shamrock-0-e2b-q4_k_m.gguf and mmproj-shamrock-0-e2b-f16.gguf for llama.cpp runtime.

Good For

  • Non-commercial Language Learning Applications: Ideal for personal projects and research focused on Japanese, Korean, or Simplified Chinese language acquisition.
  • Integration with Oxalis Ecosystem: Designed to power the language-learning features within the Oxalis desktop application.
  • Developers requiring specialized multilingual models: Suitable for those building tools that benefit from fine-tuned translation and conversational AI in the specified languages.

Finetunes
1 models