boyanbiji/Qwen3-1.7B-OPD-step120

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 25, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The boyanbiji/Qwen3-1.7B-OPD-step120 is a 2 billion parameter causal language model based on the Qwen3 architecture, developed by boyanbiji. This model is a student model trained using On-Policy Distillation (OPD) from a larger Qwen3-8B teacher model, specifically optimized for mathematical reasoning tasks. It leverages the DeepMath-103K dataset and supports a context length of up to 32768 tokens, making it suitable for complex problem-solving in mathematics.

Loading preview...

Overview

The boyanbiji/Qwen3-1.7B-OPD-step120 is a 2 billion parameter language model derived from the Qwen/Qwen3-1.7B-Base architecture. It has been specifically fine-tuned using On-Policy Distillation (OPD), a technique where a smaller "student" model learns from a larger "teacher" model. In this case, the student model was distilled from a Qwen/Qwen3-8B teacher model, focusing on non-thinking mode responses.

Key Capabilities

  • Mathematical Reasoning: The model's training on the zwhe99/DeepMath-103K dataset indicates a strong specialization in mathematical problem-solving and reasoning tasks.
  • Efficient Performance: As a 2 billion parameter model, it offers a more efficient alternative compared to larger models while retaining specialized capabilities.
  • Extended Context Window: Supports a maximum response length of 8192 tokens and an overall context length of 32768 tokens, enabling the processing of lengthy mathematical problems or complex instructions.
  • Bfloat16 Precision: Utilizes bfloat16 precision for potentially faster inference and reduced memory footprint.

Good For

  • Mathematical Applications: Ideal for tasks requiring mathematical understanding, problem-solving, and generating mathematical explanations.
  • Resource-Constrained Environments: Suitable for deployment where computational resources are limited, but specialized mathematical capabilities are needed.
  • Research in Distillation: Provides a practical example of On-Policy Distillation for improving smaller models' performance in specific domains.

Finetunes
1 models