Skip to content

Qwen2.5-Coder-3B-Instruct-GPTQ-Int8

A 3B code model tuned for generation, reasoning and repair

Visit official site

What Qwen2.5-Coder-3B-Instruct-GPTQ-Int8 is

The model is described as an instruction-tuned GPTQ Int8 release of the Qwen2.5-Coder series, trained for code generation, reasoning and repair.

What you can do with it

  • Generate source code from a prompt
  • Reason about code logic before editing
  • Detect and fix errors in code

Getting started

  • Install transformers at version 4.37.0 or newer
  • Load the model and tokenizer from the Hub
  • Apply the chat template and generate output

Pros & cons

✓ What we like

  • Code repair is listed alongside generation
  • Math and general abilities are stated as retained

! What to watch out for

  • Loading asks for transformers at version 4.37.0 or newer
  • Inputs go through the chat template before generation

FAQ

What quantisation is used?

The material states GPTQ 8-bit quantisation.

How long is the context?

The material lists a context length of 32768 tokens.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review