OFFSITE.DARK
← Greps

Nov 6, 2024

1 min

Uploaded by Qwen

HF Model
Qwen/Qwen2.5-Coder-32B-Instruct
Base Model
Qwen/Qwen2.5-Coder-32B
Pipeline
text-generation
Downloads
1,773,555
Likes
2046
  • qwen
  • code
  • codeqwen
  • text-generation
  • secure-code
  • static-analysis

oss-powerhouse

Community upload on Hugging Face — indexed for inquiry, not endorsement.

Qwen2.5-Coder-32B-Instruct — Code Model for Security Analysis

Open-source model indexed for security research inquiry — Qwen2.5-Coder-32B-Instruct with 1.7M+ downloads, backbone for pentest fine-tunes and code security tools.

What Is This?

Qwen2.5-Coder-32B-Instruct is Alibaba's open-weight code generation model on Hugging Face. With 1.7M+ downloads, it is the dominant OSS coder in the HF ecosystem — and the base for numerous security-adjacent fine-tunes (pentest GGUFs, CodeSentry agents, vulnerability research Spaces).

Indexed for inquiry. Not an OFFSITE.DARK artifact.

Metadata

FieldValue
AuthorQwen (Alibaba)
Parameters32B
Formatsafetensors
Downloads~1.77M
Likes2,046
LicenseApache 2.0

Security Research Relevance

Unlike dedicated cyber-LLMs, general coders excel at:

  • Static analysis assistance — explain dataflow, identify sink sources
  • Patch diff review — summarize security-relevant changes
  • Exploit PoC scaffolding — when guardrails permit (varies by deployment)
  • Fine-tune base — fawazo/qwen2.5-coder-3b-pentest-gguf and similar distillations

HF Spaces like YashashviAlva/codeSentry and yasserrmd/CodeCompliance demonstrate security tooling built directly on Qwen2.5-Coder.

Comparison Targets

Benchmark against:

  • Security-specific models (CodeLlama CVE LoRA, security-coding-gemma4)
  • Larger Qwen3-Coder releases as they appear
  • DeepSeek-Coder variants on identical CWE classification tasks

Attribution

Open-source model indexed for security research inquiry. Published on Hugging Face by Qwen.