ZhuofengLi/Qwen3.5-27B-hardtests-sft-synthetic-merged

VISIONConcurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 23, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

ZhuofengLi/Qwen3.5-27B-hardtests-sft-synthetic-merged is a 27 billion parameter language model based on the Qwen3.5 architecture, fine-tuned by ZhuofengLi. This model is specifically optimized for enhanced performance on challenging code and reasoning tasks. It was created by merging a LoRA adapter, which was SFT-trained on synthetic hard-tests data, into the Qwen/Qwen3.5-27B base model. Its primary strength lies in tackling complex logical and programming challenges.

Loading preview...

Model Overview

This model, ZhuofengLi/Qwen3.5-27B-hardtests-sft-synthetic-merged, is a 27 billion parameter language model derived from the Qwen3.5-27B base architecture. It has been specifically fine-tuned by ZhuofengLi to excel in demanding code and reasoning tasks.

Key Capabilities & Training

The model's enhanced capabilities stem from a LoRA adapter that was merged into the base model. This adapter underwent Supervised Fine-Tuning (SFT) using synthetic hard-tests data, specifically designed to improve performance on complex logical and programming challenges. The merge process utilized ms-swift framework (version 4.4.2) and maintained bf16 precision.

Usage

Developers can easily integrate this model using the transformers library:

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("ZhuofengLi/Qwen3.5-27B-hardtests-sft-synthetic-merged", torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("ZhuofengLi/Qwen3.5-27B-hardtests-sft-synthetic-merged")

When to Use This Model

  • Complex Code Generation: Ideal for scenarios requiring robust code solutions.
  • Advanced Reasoning Tasks: Suited for applications demanding strong logical inference and problem-solving.
  • Benchmarking Hard Problems: Useful for evaluating performance on challenging, synthetically generated test cases.