ScienceOne-AI

DeepSeek-671B-SFT-Guide. An open-source solution for full parameter fine-tuning of DeepSeek-V3/R1 671B, including complete code and scripts from training to inference, as well as some practical experiences and conclusions. (DeepSeek-V3/R1 满血版 671B 全参数微调的开源解决方案，包含从训练到推理的完整代码和脚本，以及实践中积累一些经验和结论。)

DeepLiterature. DeepLiterature: A fully open-source intelligent research assistant that integrates search, code execution, link resolution, and information expansion, with multiple tools working together to facilitate scientific research.（DeepLiterature：全流程开源的智能科研助手，集成搜索、代码执行、链接解析和信息拓展，多工具协同助力科学研究。）

AutoThink. AutoThink is a reinforcement learning framework designed to equip R1-style language models with adaptive reasoning capabilities. Instead of always thinking or never thinking, the model learns when to engage in explicit reasoning, balancing performance and efficiency.

S1-DeepResearch. Python

S1-Agent.

S1-Parser. Python

Agent-Interaction-Protocol. Agent Interaction Protocol is a distributed agent interaction protocol. It defines communication mechanisms for multi-agent collaboration, multi-tool invocation, and multi-modal data access in scientific scenarios.

PDE-Agent. A toolchain-augmented multi-agent framework for PDE solving

MixReason. 让大模型在思考过程中使用搜索和代码工具

ScienceEval. An open-source evaluation suite designed for the ScienceOne Base models. The project is actively maintained, and we welcome community contributions.

S1-Omni-Image. A Unified Multimodal Model for Scientific Image Understanding and Generation

S1-MatAgent. The official repository of S1-MatAgent.

S1-Base-API. 磐石科学基础大模型API

PhysLogic. Python

Code-Reasoner. Multimodal Physics Agent with Code Enhancement: 2nd place in ICML 2025 AI4MATH Challenge

ScienceOne-AI.github.io. under construction