InternLM 开源 JanusCoder：面向多模态代码智能的视觉-程序统一接口

InternLM; JanusCoder

Repository navigation

NotificationsYou must be signed in to change notification settings; Fork8 (8); Star81 (81)You must be signed in to star a repository

Repository files navigation

README; Apache-2.0 license

🗞️ Updates

2026-05-09: We are releasing our data at 🤗JanusCode-800K; 2026-01-26: JanusCoder is accepted by ICLR 2026! 🎉 We will release more resources soon!; 2025-11-15: We release data samples of JanusCode-800K, check section Data Samples for more details; 2025-10-29: We release the preprint and checkpoints of JanusCoder series 🤗; 2025-10-07: Initial release of our technical report, code, data samples, and 🌐 Project Page. Check it out! 🚀

Note

This release represents the public implementation; the full implementation and data will be made available after internal company policy requirements are met.

📑 Intro

JanusCoder is a suite of open models that establish a unified visual–programmatic interface for multimodal code intelligence. The models (JanusCoder and JanusCoderV) handle text-centric and vision-centric tasks in a single framework—from chart-to-code and web UI generation/editing to dynamic theorem visualizations—and show strong results across public benchmarks, approaching or even surpassing proprietary systems.

Note

Due to company policy, we need some additional time to release all datasets and checkpoints. If you require access to more data, please feel free to send qiushisun@connect.hku.hk an email.

🤗 Checkpoints

We provide the checkpoints of JanusCoder series on Hugging Face:

⚙️ Data Synthesis Toolkit

We provide a versatile data synthesis toolkit that generates multimodal code data across heterogeneous domains—ranging from charts and Web UIs to visual artifacts and code-driven animations—while greatly reducing engineering efforts for large-scale corpus creation. 

Since the process of building JanusCode data involves a variety of synthesis pipelines, we provide a few examples here:

Example Workflows:

Extend, refine and derive new text-centric data for chart tasks

Extend and derive new text-centric data for visual editing tasks

More scripts will be released soon.

Data Samples:

We provide text-centric data samples at this link; We provide vision-centric data samples at this link

🧪 Training

We primarily follow the official training pipelines provided. Users can directly refer to the linked repositories for detailed instructions on SFT.

We also provide some typical training configuration file for llamafactory users in training_files.

All our experiments were conducted on interconnected 8× H800 GPUs.

📏 Evaluation

We provide several ready-to-use scripts to quickly reproduce our experimental results. You can replace them with other scripts under the evaluation directory to evaluate different tasks, for example:

For evaluations on LiveCodeBench-v6, MBPP+: We directly adopt the evaluation scripts provided by OpenCompass.

For evaluations on Artifactbench and Interactscience can be run the instructions in their original repositories. You can fork/clone the original repos to reuse their official environments and configs.

💽 Data Preview

Here we provide data samples of the various data types covered by JanusCode-800K:

text-centric data: [Link]; vision-centric data: [Link]

the full implementation and data will be made available after internal company policy requirements are met, stay tuned!

📚License

This project is licensed under the Apache 2.0 License. See the LICENSE file for details.

💐Acknowledgement

We thank Fangzhi Xu and Zhangyue Yin for their valuable feedback on improving the manuscript's presentation, and Zichen Ding for the assistance in resolving issues encountered during the training of InternVL 3.5. We are also grateful to Fahd Mirza for creating an nice YouTube video on running JanusCoder locally.

📋 Citation

🫶 If you are interested in our work or find the repository / checkpoints / benchmark / data helpful, please consider using the following citation format when referencing our papers: