Surpem/Supertron2-1.7B

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:2BQuant:BF16Ctx Length:32kPublished:May 14, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

Supertron2-1.7B by Surpem is a 1.7 billion parameter instruction-tuned causal language model built upon Qwen3-1.7B, designed for efficient daily use. It excels in multi-step reasoning, mathematical problem-solving, and code generation across various languages. This compact model provides strong performance in science, general knowledge, and conversation, making it suitable for consumer hardware.

Loading preview...

Supertron2-1.7B: A Compact, Efficient Instruction-Tuned Model

Supertron2-1.7B, developed by Surpem, is a 1.7 billion parameter instruction-tuned causal language model based on Qwen3-1.7B. Licensed under Apache 2.0, it is engineered to be a reliable and efficient daily driver, offering robust performance across a variety of domains while remaining lightweight enough for consumer hardware.

Key Capabilities

  • Reasoning: Designed for clear, multi-step reasoning, capable of methodically breaking down complex problems.
  • Math: Handles arithmetic, algebra, word problems, and structured problem-solving, providing explanations and concise answers.
  • Coding: Writes, debugs, and explains code in popular languages like Python, JavaScript, and C++, understanding syntax and algorithmic reasoning.
  • Science & General Knowledge: Broad instruction tuning enables technical conversations, clear explanations of difficult concepts, and assistance with research and analysis.
  • Instruction Following: Responsive to natural language instructions, adapting to desired formats and tones without complex prompting.

Good For

  • Developers seeking a compact yet powerful model for diverse tasks.
  • Applications requiring strong reasoning and problem-solving abilities.
  • Code generation, debugging, and explanation.
  • General conversational AI and knowledge-based assistance.
  • Deployment on consumer-grade hardware with minimal VRAM requirements (e.g., 5GB for bfloat16, 3GB for 4-bit quantized).