harithoppil/Qwen3-0.6B-English

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:0.8BQuant:BF16Ctx Length:32kPublished:Feb 16, 2026License:apache-2.0Architecture:Transformer Open Weights Warm

harithoppil/Qwen3-0.6B-English is a specialized 0.5 billion parameter causal language model, derived from Qwen/Qwen3-0.6B, with a context length of 32,768 tokens. This version has been vocabulary-pruned to focus exclusively on English text, programming code, mathematics (LaTeX), and logical reasoning by removing non-English tokens. It features Qwen3's unique ability to seamlessly switch between 'thinking' and 'non-thinking' modes, optimizing performance for complex logical tasks and general dialogue respectively.

Loading preview...

Qwen3-0.6B-English: English-Centric Optimization

This model is a specialized, vocabulary-pruned version of the Qwen3-0.6B series, developed by harithoppil. It reduces the vocabulary from 151,643 to 101,816 tokens by removing non-English characters (like CJK, Cyrillic, Arabic) while retaining full support for English, Programming Code, Mathematics (LaTeX), and logical reasoning. This optimization results in a smaller embedding layer and LM head, leading to reduced parameter count (~0.5B) and slightly faster logits calculation.

Key Capabilities

  • Dual-Mode Operation: Uniquely supports seamless switching between a 'thinking mode' for complex logical reasoning, math, and coding, and a 'non-thinking mode' for efficient, general-purpose dialogue.
  • Enhanced Reasoning: Demonstrates significant improvements in mathematics, code generation, and commonsense logical reasoning, particularly in its thinking mode.
  • Human Preference Alignment: Excels in creative writing, role-playing, multi-turn dialogues, and instruction following.
  • Agent Capabilities: Features strong tool-calling abilities, enabling precise integration with external tools for complex agent-based tasks.

Use Cases

This model is ideal for applications requiring high performance in English-centric tasks, code generation, mathematical problem-solving, and logical reasoning. Its dual-mode functionality allows developers to optimize for either deep analytical processing or efficient conversational interactions, making it versatile for various English-language AI applications.

Finetunes
2 models