Qwopus3.8-27B-Coder-Flash

Qwopus3.8-27B-Coder-Flash

Writes code, calls tools and works through software tasks as a coding agent. A 27.8B fine-tune of Qwen3.8-27B by Jackrong, trained on about 200,000 code and reasoning examples, for developers with a strong GPU.

Made by Jackrong on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: Jackrong built this model from Qwen3.8-27B and tuned it to write code and act as a software agent. In an agent loop it reads a task, edits code, checks what its tools return, runs tests and repairs its own mistakes. Compact reasoning is a design goal, though the card does not measure it.

Who it suits: Developers and agent builders who want a large open coding model they can run themselves and who have a strong GPU or lots of memory. The files carry the Apache 2.0 license.

Industry | Who uses it | How they use it | Example
Software | Developers | Generate and refactor code with a local model | Asking for a new function with edge cases handled
Developer tools | Agent builders | Power a coding agent that calls tools and tests its work | A local agent that edits files and runs tests
Creative coding | Hobbyists | Build visual scenes in code | Modeling a 3D seaside town, as in the card's showcase

Anyone who wants chat, summaries or open question answering can skip it, since the card names those as outside its focus.

Where to use it:

• Python with the transformers library, the one Hugging Face lists
• A machine with plenty of memory, since the weights are large
• Windows, macOS or Linux with Python, for downloading the files

Use case | Who | What they get back
Code generation | Developers | Code that handles edge cases and follows the instructions
Debugging | Programmers | Fixes proposed after reading tool and test output
Agent work | Agent builders | Multi-turn tool calls with error recovery

How to set it up:

• Python and the huggingface_hub package
• The transformers library to load the model
• Room for the weights, about 55.6 GB in BF16, or about 13.9 GB at 4-bit as an estimate

1. Install Python if you do not have it yet.
2. Install the huggingface_hub package with the first command below.
3. Download the model files with the second command.
4. Load the model with the transformers library, following the model page.

pip install -U huggingface_hub
hf download Jackrong/Qwopus3.8-27B-Coder-Flash

How to use it: Once the files are on your computer, you load the model and give it a coding task, or run it inside an agent that can call tools. The card text available here shows no run command, so use the model page for the exact steps.

Example: You give an agent running this model a bug report and a test suite. It reads the code, makes a change, looks at the test output, fixes any error and hands back the finished change.

For newcomers and experts: A good first step is to download the files and read the model page. Then ask for one small function and run your own tests on the result.

The training mix has three parts. About 20K agent examples came mostly from Qwen3.8-Flash-Max. Code examples came from teacher models in the Qwen3.8 family plus DeepSeek-V4-Flash. Reasoning and instruction data came from the NVIDIA Nemotron series. The card says it does not measure any cut in reasoning length.

Teacher models write agent and code examples
Fine-tune Qwen3.8-27B on about 200,000 examples
Load the checkpoint
Give a coding task or agent loop
Check the result with your own tests

Good to know:

• License: Apache 2.0
• Cost to run: none named
• Data it reaches: the download fetches files from Hugging Face, and the source names nothing else
• Languages listed: English and Chinese
• Hugging Face lists 27.8B parameters and 22 quantized versions made by other people
• Do not use it for fully automated decisions with legal, medical or financial effects on people, without human review

Title: Training examples, in thousands
All examples: 200 K
Agent interaction: 20 K

Title: Weights size in GB
BF16: 55.6 GB
4-bit estimate: 13.9 GB

On Hugging Face, 13 November 2026

• Task: image text to text
• Library: transformers
• Parameters: 27.8B
• Weights: about 55.6 GB in BF16; about 13.9 GB at 4-bit (estimate)
• Formats: Safetensors
• Downloads in the last 30 days: 1,496
• Likes: 34
• License: apache-2.0
• Languages: en, zh
• Based on: unsloth/Qwen3.8-27B
• Quantized versions: 23
• Tags: transformers, safetensors, qwen3_5, image-text-to-text, vision, multimodal, conversational, qwopus3_8, vllm, fine-tuned, coding, code-generation

Links

• Model: https://huggingface.co/Jackrong/Qwopus3.8-27B-Coder-Flash
• Files: https://huggingface.co/Jackrong/Qwopus3.8-27B-Coder-Flash/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download Jackrong/Qwopus3.8-27B-Coder-Flash

Then follow the usage in its model card. Nothing in this kit runs by itself.