Qwen3.8-27B-opencode-adapted for Apple Silicon (MLX, 4-bit)

harness-hub/Qwen3.8-27B-opencode-adapted (revision b4171c0) merged into Qwen/Qwen3.8-27B (revision 3756181) and converted to MLX with mlx-vlm, vision encoder and Qwen's MTP layer included. In this build the linear layers are quantized to 4-bit (affine, group 64). Size: 16.3 GB.

harness-hub trained this LoRA with reinforcement learning (AsyncGRPO in TRL) to be the model behind the OpenCode coding agent on software-engineering and terminal tasks. Their published results, under OpenCode with thinking off: SWE-bench Pro (hard 51) 35.3% โ†’ 51.0% and Terminal-Bench 2.1 46.4% โ†’ 52.4% against the base model, and with thinking off it matches the base model thinking at its highest effort while generating about a sixth of the tokens. See the original adapter card for the method, data and full results.

Build Size Same next token as bf16 KL vs bf16
netprojects/Qwen3.8-27B-opencode-adapted-MLX-6bit (recommended) 23.1 GB 98.7% 0.0014
netprojects/Qwen3.8-27B-opencode-adapted-MLX-8bit 30.0 GB 99.2% 0.0004
netprojects/Qwen3.8-27B-opencode-adapted-MLX-4bit 16.3 GB 96.7% 0.0203
netprojects/Qwen3.8-27B-opencode-adapted-MLX-bf16 55.6 GB reference reference
netprojects/Qwen3.8-27B-opencode-adapted-MLX-adapter 638 MB LoRA only, for mlx-lm on any Qwen3.8-27B MLX build

The 6-bit build is the recommended balance of size and fidelity; 8-bit is closest to bf16, 4-bit is the smallest.

Use with OpenCode

The adapter was trained and evaluated with thinking off, so keep it off. Serve the model with mlx-lm's OpenAI-compatible server:

pip install -U "mlx-lm>=0.32.2"
mlx_lm.server --model netprojects/Qwen3.8-27B-opencode-adapted-MLX-4bit --port 8000 \
  --chat-template-args '{"enable_thinking": false}'

Add it to ~/.config/opencode/opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "local": {
      "npm": "@ai-sdk/openai-compatible",
      "options": { "baseURL": "http://localhost:8000/v1", "apiKey": "dummy" },
      "models": { "netprojects/Qwen3.8-27B-opencode-adapted-MLX-4bit": {} }
    }
  }
}
opencode run --model local/netprojects/Qwen3.8-27B-opencode-adapted-MLX-4bit "Find out why the tests in tests/test_parser.py fail and fix the bug."

Use from Python

Text only, with mlx-lm:

from mlx_lm import load, generate

model, tokenizer = load("netprojects/Qwen3.8-27B-opencode-adapted-MLX-4bit")
messages = [{"role": "user", "content": "Write a Python function that parses '1.5M' into bytes."}]
prompt = tokenizer.apply_chat_template(messages, add_generation_prompt=True, enable_thinking=False, tokenize=False)
print(generate(model, tokenizer, prompt=prompt, max_tokens=512))

With images, with mlx-vlm:

pip install -U "mlx-vlm>=0.7.11"
mlx_vlm.generate --model netprojects/Qwen3.8-27B-opencode-adapted-MLX-4bit --image screenshot.png \
  --prompt "What error does this terminal show?" --max-tokens 300   # thinking stays off unless --enable-thinking

The tokenizer and chat template are harness-hub's files, unchanged (same vocabulary and tool-call format as Qwen3.8-27B). Tool calls use Qwen's XML format (<tool_call><function=...><parameter=...>), which mlx-lm's server returns as OpenAI tool_calls.

Faster decoding: MTP and DSpine

This build keeps Qwen3.8-27B's own multi-token-prediction (MTP) layer in model-mtp.safetensors, quantized to match the build. The adapter does not change that layer, so it is Qwen's original MTP head. mlx-vlm uses it as a built-in draft model when you pass the same repository as --draft-model:

mlx_vlm.generate --model netprojects/Qwen3.8-27B-opencode-adapted-MLX-4bit --draft-model netprojects/Qwen3.8-27B-opencode-adapted-MLX-4bit \
  --draft-kind mtp --temperature 0 --max-tokens 512 --prompt "Write a Python function that checks whether a number is prime."

On this build MTP makes decoding about 1.9ร— faster with no extra download. The DSpine drafter with dspine-hybrid-mlx also works with these builds. Our measurements (greedy, thinking off, 5 prompts ร— 256 tokens from GSM8K, MATH-500, HumanEval, MBPP and MT-Bench, with other services running on the Mac, so treat them as indicative):

Build Plain MTP (mlx-vlm 0.7.11) DSpine v0.2.1 (4-bit drafter, block 5)
4-bit 33 tok/s 62 tok/s (ร—1.91) 73 tok/s (ร—2.29)
6-bit 20 tok/s 21 tok/s (ร—1.04) 60 tok/s (ร—2.72)
8-bit 17 tok/s 37 tok/s (ร—2.15) 54 tok/s (ร—3.03)

Both methods check every drafted token with the full model. With greedy decoding the text matches plain decoding until a near-tie, where the batched check can round the other way, so long answers can diverge slightly.

What was checked

On a MacBook Pro M5 Max (128 GB), 18 November 2026, with mlx 0.32.4, mlx-lm 0.32.2 and mlx-vlm 0.7.11.

  • Merge. All 256 adapted modules (q, k, v, o in the 16 full-attention layers; gate, up, down in all 64 layers) equal W + 2.0 ยท B @ A rounded to bf16, checked element by element against an independent float64 computation; every other tensor is byte-identical to Qwen3.8-27B. The adapter does not touch the vision encoder or the Gated DeltaNet layers.
  • Quantization. Each quantized build is checked to be exactly the affine (group 64) quantization of the bf16 build.
  • Agreement with bf16 (12 short prompts, thinking off, 949 teacher-forced positions): the 8-bit build picks the same next token as bf16 on 99.2% of positions, 6-bit on 98.7% and 4-bit on 96.7% (mean KL 0.0004 / 0.0014 / 0.0203).
  • Tool calling with the included chat template, thinking off: one agent-style prompt with bash and read tools; 8-bit, 6-bit, 4-bit each produced a well-formed bash tool call in Qwen's XML format.
  • Coding. Three short Python tasks run against asserts: 8-bit 3/3, 6-bit 3/3, 4-bit 3/3.
  • Vision (mlx-vlm): asked to name the shapes and colours in a test image, 8-bit, 6-bit, 4-bit answered correctly, and answered a text-only question in the same session.
  • MTP layer. The bundled MTP tensors equal Qwen3.8-27B's mtp.* weights (norms in mlx-vlm's stored form), quantized exactly like the rest of the build; each build loads normally in mlx-lm and mlx-vlm with the extra file present.
  • Source files. Every file of both pinned upstream revisions matches the Hub's SHA-256 hashes.

Credits

The adapter is by harness-hub; the base model is Qwen3.8-27B by the Qwen Team, Alibaba Cloud. If you use the model, please cite their work (see the adapter card and the Qwen3.8-27B card).

License

Apache-2.0, as Qwen3.8-27B and the harness-hub adapter. LICENSE is the Apache-2.0 text from the Qwen3.8-27B repository and NOTICE describes the conversion and credits harness-hub and the Qwen Team (Alibaba Cloud).

Published by NetProjects (Internet Projects Ltd). Not affiliated with Hugging Face, harness-hub or Alibaba.

More MLX builds from NetProjects: huggingface.co/netprojects ยท netprojects.com

Downloads last month
-
Safetensors
Model size
28B params
Tensor type
U32
ยท
BF16
ยท
MLX
Hardware compatibility
Log In to add your hardware

4-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for netprojects/Qwen3.8-27B-opencode-adapted-MLX-4bit

Base model

Qwen/Qwen3.8-27B
Quantized
(6)
this model

Collection including netprojects/Qwen3.8-27B-opencode-adapted-MLX-4bit