gpt_academic

A Python project that puts a practical interface in front of GPT, GLM, and other large language models, tuned for reading, polishing, and writing papers, with function plugins and custom shortcut buttons.

GPT Academic describes itself as a practical interactive interface for GPT, GLM, and other large language models, with particular attention to paper reading, polishing, and writing. It is modular, with custom shortcut buttons and function plugins, supports analysis of Python and C++ projects, and can query multiple models in parallel. The project is Python under GPL-3.0, and the README is written mostly in Chinese.

The top of the README is a dated news list. A new GUI frontend is in testing as of January 2026. The Dockerfile build efficiency was improved in August 2025, and a video shows a three minute setup for qwen2.5-max. Custom fonts arrived in February 2025, the whl file server was restored after a power outage in October 2024, and Doc2x PDF paper translation was added in May 2024. Full support for Chinese models like Qwen, GLM, and DeepseekCoder was announced in March 2024.

The academic angle shows in the plugin list. One plugin reads and translates latex or PDF papers and generates a summary. Another translates or polishes latex documents, and a batch comment plugin generates function comments. An arXiv assistant translates abstracts and downloads PDFs, and a latex proofreading plugin checks grammar and spelling Grammarly style, outputting a comparison PDF. A Google Scholar helper writes related works from a search page URL.

Model support is broad. The README lists Baidu Qianfan and Wenxin Yiyan, Qwen, Shanghai AI-Lab, Xunfei Spark, LLaMa2, GLM4, DALLE3, and DeepseekCoder, plus local models like ChatGLM2 and MOSS. Multiple API keys can coexist in a single configuration. Installation notes describe a config private.py file that overrides config.py, environment variables with a documented reading priority, and deployment options from a Windows one click script to Sealos, WSL2, and running under a subpath with FastAPI.

Not everything is text based. Mermaid rendering lets the models draw flowcharts, state transition diagrams, Gantt charts, and pie charts. Real time voice conversation listens to audio and finds its own moment to answer, and a void terminal plugin dispatches other plugins using natural language. There is internet information aggregation and image generation support, and the version history runs from 1.0 up to a planned 3.80. Each file is documented in a self analysis report that the project can regenerate.

The README reads like a changelog plus a features list: recent work on the master branch includes a new GUI in testing and faster Docker builds, and the plugin system keeps growing.

• Source 1 · Project repository
• Source 2 · README