Model Releases

First unified scientific large model LOGOS officially open-sourced

AI summary

LOGOS, open-sourced by ATH-Token Foundry in collaboration with Renmin University of China's Gaoling School of Artificial Intelligence, is the first multi-domain scientific generative foundation model based on a unified "scientific grammar." LOGOS-1B (1B parameters) matches or surpasses domain-specific methods on six major scientific tasks: its pure-sequence paradigm for pocket-conditioned ligand generation surpasses 3D diffusion models for the first time, outperforming NatureLM (8×7B); retrosynthesis prediction achieves 74.8% Top-1 accuracy; pocket site identification reaches 58.5% Top-n accuracy using sequence alone; and MOF material generation improves NBB to 17.78%. The model uses a unified vocabulary to encode proteins, small molecules, and other entities as discrete tokens, and achieves sequence prediction without 3D coordinates through spatial interaction discretization. Pretraining and downstream tasks share consistent formats and objectives, and cross-domain knowledge transfer has been experimentally verified as effective. Model weights, inference code, and the technical report have been fully open-sourced.

Key Takeaways

  • LOGOS,open-sourced by ATH-Token Foundry in collaboration with Renmin University of China's Gaoling…
  • " LOGOS-1B (1B parameters) matches or surpasses domain-specific methods on six major scientific…
  • retrosynthesis prediction achieves 74.
  • 8% Top-1 accuracy
  • pocket site identification reaches 58.

Quick Facts

Published
Updated
Source
公众号:通义实验室(千问)
Category
Model Releases
AI score
75
Read the original Read the full article at the original source
Why it matters

LOGOS用一套统一科学语法把蛋白质、小分子、材料等塞进同一个LLM框架,纯序列建模就干过了专用扩散模型,参数量却只有NatureLM的1/56,做AI4S的得认真看一眼。

Source & copyright

This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.

Original article · AIHOT entry

How to Cite

When citing this article, keep the title, date, and permalink, and credit Orange News as the source.

Orange News: First unified scientific large model LOGOS officially open-sourced. June 18, 2026. https://new.nat6.net/en/news/cmqiw7jw604v7sl5w10zed5s8/

View all →