#2626th of 80 in Open modelsFor data and documents, data analysis
deepanalyze
8B agentic LLM that runs data science tasks and writes analyst-style reports
- Stars
- 4.8k
- License
- MIT
- Last commit
- Nov 2026
- Language
- Python
3,971 on 22 Apr 20264,809 on 11 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
DeepAnalyze is an 8B model (DeepAnalyze-8B, on Hugging Face and ModelScope) trained to plan, write and execute code for data preparation, analysis, modeling, visualization and report generation. It takes CSV, Excel, JSON, XML, YAML, TXT, Markdown and database sources and produces a research report. The repo includes a vLLM-based Python API, an HTTP API server, two web UIs, a Jupyter UI, a CLI, training code and the 500K-sample training dataset.
Who it is for: Data teams and researchers wanting a self-hosted autonomous data-analysis model
Strengths
- Weights, training data (DataScience-Instruct-500K) and training code are all published
- Handles structured, semi-structured and unstructured inputs in one open-ended run
- README gives vLLM settings per GPU size, including 4-bit and 8-bit variants
- Multiple frontends: web UI, Jupyter, CLI, and a Python API
Weaknesses
- No Docker or compose setup for the main demo; needs npm, vLLM and shell scripts
- README labels the demo UI an initial version
- Needs a GPU of 16GB or more for local inference, with reduced context on smaller cards
- Quick-start example uses hardcoded author-specific paths that must be replaced
What it needs
- GPU required
- Docker + Compose
- Needs vLLM >=0.8.5, PyTorch, Transformers, Node.js/npm (web UI), Python 3.12
- Models: DeepAnalyze-8B, 4-bit quantized, 8-bit quantized, OpenAI-style API endpoints
- port 4000
Also in Open models
See all 80| Rank | Project | Score |
|---|---|---|
| 1 | ruviewWiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose | 94 out of 100 |
| 2 | layaSingle-pass multilingual classifier returning typed decisions with calibrated odds | 88 out of 100 |
| 3 | rf-detrReal-time transformer models for object detection, segmentation and keypoints | 81 out of 100 |
| #4 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 78 out of 100 |
| #5 | open_clipPyTorch library for training and running CLIP-style image-text models | 78 out of 100 |