File size: 2,814 Bytes
d65ad3b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 | ---
license: apache-2.0
base_model:
- Qwen/Qwen3.6-27B
library_name: mlx
tags:
- mlx
- qwen3.6
- qwen3_5
- apple-silicon
- tool-calling
- go
- golang
---
# Qwen3.6-27B-go-v1 MTP BF16 MLX
A Go-focused Qwen3.6-27B model for Apple Silicon, packaged in MLX.
Use it as a coding assistant for Go projects: generating focused patches, explaining diffs, tightening tests, reading command output, and making small repo-aware edits. It is tuned for tool-calling workflows where the assistant has to inspect files, run commands, and keep changes narrow.
This is the native-MTP BF16 full-precision variant. This native-MTP package includes the model's MTP weights for runtimes that support the Qwen3.6 dense MTP layout, such as oMLX. If a loader rejects MTP weights, use the matching plain repo instead.
## Which Variant Should I Use?
- **Use this repo** if your runtime supports native MTP for dense Qwen3.6 MLX models and you want the native-MTP BF16 full-precision package.
- Use [`jedisct1/Qwen3.6-27B-go-v1-MTP-4bit.mlx`](https://huggingface.co/jedisct1/Qwen3.6-27B-go-v1-MTP-4bit.mlx) for the native-MTP 4-bit OptIQ sibling.
- Use [`jedisct1/Qwen3.6-27B-go-v1-MTP-8bit.mlx`](https://huggingface.co/jedisct1/Qwen3.6-27B-go-v1-MTP-8bit.mlx) for the native-MTP 8-bit sibling.
- Use [`jedisct1/Qwen3.6-27B-go-v1-bf16.mlx`](https://huggingface.co/jedisct1/Qwen3.6-27B-go-v1-bf16.mlx) if your runtime does not support native MTP weights.
## Usage
Requires [mlx-lm](https://github.com/ml-explore/mlx-lm):
```bash
pip install mlx-lm
```
```python
from mlx_lm import load, generate
model, tokenizer = load("jedisct1/Qwen3.6-27B-go-v1-MTP-bf16.mlx")
messages = [
{"role": "system", "content": "You are an expert Go developer."},
{"role": "user", "content": "Generate a focused patch that replaces panic() in loadConfig() with returned errors and table-driven tests."},
]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
response = generate(model, tokenizer, prompt=prompt, max_tokens=500)
print(response)
```
## What It Is Good At
- Writing idiomatic Go patches from a concise change request.
- Explaining Go diffs in commit-message style.
- Following tool-calling workflows where it needs to inspect files before editing.
- Keeping changes focused instead of turning small fixes into broad rewrites.
- Working with tests, compiler errors, command output, and repository context.
## Limitations
- Outputs should be reviewed before use, especially concurrency code, unsafe/cgo code, and changes that affect security boundaries.
- The model works best on focused Go changes, tests, and explanations. Very large refactors may need to be split into smaller steps.
- Tool calling depends on the runtime and client preserving the chat template and tool schema format.
|