README.md

---
license: apache-2.0
base_model: microsoft/AesCode-8B
tags:
  - mlx
  - mlx-vlm
  - vision
  - multimodal
  - image-text-to-text
  - code-generation
  - html
  - design
  - reinforcement-learning
pipeline_tag: image-text-to-text
library_name: mlx-vlm
---

# AesCode-8B — Vision + Code (bf16, MLX)

Full-precision **bf16** MLX conversion of Microsoft's **AesCode-8B**, with the **vision tower
kept intact** so it reads a design or screenshot and writes the HTML/CSS back on Apple
Silicon. No quantization: every weight is the original bf16. **750 tensors**, 17.5 GB, four
shards.

**This is the full-quality build. Smaller sizes that passed the same vision check:
[8-bit (about half the size, for 16 GB machines)](https://huggingface.co/divinetribe/AesCode-8B-MLX-8bit)
· [4-bit (smallest, for 8 GB machines)](https://huggingface.co/divinetribe/AesCode-8B-MLX-4bit).**

AesCode generates information-rich visual artifacts (slides, posters, dashboards, reports)
as structured, editable, verifiable HTML/CSS. It starts from
[Qwen3-VL-8B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct) and is trained with
cold-start SFT followed by GDPO across seven reward channels, separating semantic
requirements from visual cues through graph-structured supervision and decoupled cross-modal
rewards.

## What it is

- **Base:** [`Qwen/Qwen3-VL-8B-Instruct`](https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct) (Qwen team).
- **Source checkpoint:** [`microsoft/AesCode-8B`](https://huggingface.co/microsoft/AesCode-8B) (Apache-2.0).
- **Architecture:** `Qwen3VLForConditionalGeneration` — a `qwen3_vl` vision tower feeding the
  `qwen3_vl_text` decoder. 8.8B parameters.
- **This repo:** bf16 MLX weights, vision kept, for [`mlx-vlm`](https://github.com/Blaizzy/mlx-vlm).
- **Other sizes:** [8-bit](https://huggingface.co/divinetribe/AesCode-8B-MLX-8bit) · [4-bit](https://huggingface.co/divinetribe/AesCode-8B-MLX-4bit).

## Running it (MLX)

```python
from mlx_vlm import load, generate
from mlx_vlm.prompt_utils import apply_chat_template

model, processor = load("divinetribe/AesCode-8B-MLX-bf16")
prompt = apply_chat_template(processor, model.config,
                             "Rebuild this screenshot as a single HTML file.", num_images=1)
print(generate(model, processor, prompt, image=["design.png"], max_tokens=1024))
```

Text-only works with `num_images=0` and no `image=` argument.

## Worked vision example — pose-match QC

A real frame from a motion game (a translucent target avatar drawn over a live camera
scene) was fed to this converted bf16 model, asking whether the drawn figure is a correct
human pose and whether any limbs are extra or wrong. The model's unedited on-device answer:

> The translucent avatar shows a standing human pose. Arms are bent at the sides, elbows
> slightly raised, hands near the waist. Legs are straight with knees slightly bent, feet
> flat on the ground. The torso is upright, and the head is centered. No extra limbs or
> reversed joints are visible. Overall, the pose is anatomically plausible for a neutral
> standing position.

This is grounded in the actual pixels: it reads the translucency of the overlay, the single
standing figure, the bent elbows and bent knees, feet flat and head centered, and correctly
answers the core QC question — no extra limbs and no reversed joints. It placed the hands at
the waist rather than up in a full guard, so it is a genuine, imperfect QC read, shown
unedited. The same vision path that reads this frame also reads a design or screenshot and
writes **alt-text** and the **HTML/CSS** back, which is what AesCode is built for.

## Benchmarks

The single most reputable benchmark that genuinely applies is **AesCode's own evaluation**
from the paper (*AesCode: Aesthetic Code Generation with Decoupled Cross-Modal Rewards*,
[arXiv:2610.18860](https://arxiv.org/abs/2610.18860)): 300 infographic prompts scored on a
programmatic `Rule` track (Text, Bound, Chart) and a `Visual` checklist (Content, Layout,
Style), with `Overall` the mean of the two, averaged over three generations with no
selection. AesCode-8B reaches **Overall 82.94** and **Rule 90.07**, improving Visual over its
reference-conditioned Qwen3-VL-8B backbone by **31.1 points**; the paper reports this 8B
result exceeding the reference-conditioned GPT-5.5 (81.28) and Claude Opus 4.8 (80.39) rows
in the same table. All numbers are quoted from the AesCode paper and model card, not invented.

## Credits

Enormous thanks to the people whose work this conversion simply repackages:

- **The Microsoft AesCode authors** — Xin Gao, Yifan Yang, Xin Li, Pan Wang, Yan Li, Muzhao
  Tian, Mingxi Cheng, Bei Liu, Kai Qiu, Qi Dai, Dong Chen, Yue Dong, Leandros Tassiulas, Rex
  Ying, Ji Li, and Chong Luo — who built and released AesCode-8B under Apache-2.0:
  [model](https://huggingface.co/microsoft/AesCode-8B) ·
  [paper](https://arxiv.org/abs/2610.18860) ·
  [project page](https://microsoft.github.io/AesCode/) ·
  [code](https://github.com/microsoft/AesCode).
- **The Qwen team** at Alibaba — for the
  [Qwen3-VL-8B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct) base and the
  Qwen3-VL vision-language architecture.
- **Prince Canuma and the [mlx-vlm](https://github.com/Blaizzy/mlx-vlm) project**, and the
  **Apple MLX team**, for the tooling that makes vision-language models run natively on
  Apple Silicon.

The conversion to bf16 MLX is the only work done here. All model quality is theirs.

---
<!-- ccl-frontdoor -->
## Part of Claude Code Local

This model is one of the fighters in **[Claude Code Local](https://github.com/nicedreamzapp/claude-code-local)** (3.2k★), which runs Claude Code 100% on-device on Apple Silicon through an MLX-native Anthropic-API server. Not sure which local model to run as an agent? Check the **[Agent-12 local agent leaderboard](https://nicedreamzapp.github.io/agent12/)**: real agent tasks, judged by the filesystem, same hardware for every row.

Built by [Matt Macosko](https://github.com/nicedreamzapp) in Humboldt, CA. Open to work on local-AI and Apple Silicon inference: matt@ineedhemp.com.