cs.LGMar 26, 2026

Intern-S1-Pro: Scientific Multimodal Foundation Model at Trillion Scale

Authors: Yicheng Zou, Dongsheng Zhu, Lin Zhu, Tong Zhu, Yunhua Zhou, Peiheng Zhou, Xinyu Zhou, Dongzhan Zhou, 

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

We introduce Intern-S1-Pro, the first one-trillion-parameter scientific multimodal foundation model. Scaling to this unprecedented size, the model delivers a comprehensive enhancement across both general and scientific domains. Beyond stronger reasoning and image-text understanding capabilities, its intelligence is augmented with advanced agent capabilities. Simultaneously, its scientific expertise has been vastly expanded to master over 100 specialized tasks across critical science fields, including chemistry, materials, life sciences, and earth sciences. Achieving this massive scale is made possible by the robust infrastructure support of XTuner and LMDeploy, which facilitates highly efficient Reinforcement Learning (RL) training at the 1-trillion parameter level while ensuring strict precision consistency between training and inference. By seamlessly integrating these advancements, Intern-S1-Pro further fortifies the fusion of general and specialized intelligence, working as a Specializable Generalist, demonstrating its position in the top tier of open-source models for general capabilities, while outperforming proprietary models in the depth of specialized scientific tasks.

Figures & tables

Figure 1: The SAGE (Synergistic Architecture for Generalizable Experts, including three layers, Foundation, Fusion, and Evolution) framework used in Intern-S1-Pro development, illustrating the core capabilities and the integrated learning process that enables synergistic improvements across domains.
Figure 2: Left: Illustration of the expert expansion process from Intern-S1 to Intern-S1-Pro. The grouped routing strategy ensures well-trained Top-1/Top-2 experts are distributed across groups to maintain training stability. Right: The illustration of Straight-Through Estimator for enabling updating all router embeddings in every pass.
Figure 3: Training with Grouped Router can achieve absolute load balancing across devices for MoE models with a Top-k configuration of k=8 under the EP8 training strategy.
Figure 4: FoPE models each dimension as a Fourier series of different frequency components, thereby separating information more effectively and mitigating spectral damage. Inadequately trained frequency components are also clipped as their harmful influence on spectral.
Figure 5: Architecture and subsampling mechanism of the time series module.
Figure 6: The comparison of natural caption (often occurs in scientific literature) and the desired dense caption for training VLM. The key is the text should explicitly refer the visual elements.
Figure 7: The workflow of caption pipeline used in data preparation, illustrating how high-quality aligned scientific multimodal data is produced in a efficient way.
Figure 8: Comparison between FP8 mixed-precision RL training and BF16 training on a 30B MoE model. The curves show the average accuracy across multiple validation benchmarks during RL optimization. FP8 training closely matches the BF16 baseline throughout training.
 ThinkingNon-Thinking
max tokens6553632768
temperature0.80
top_p0.951.0
top_k501
Table 1: Intern-S1-Pro Evaluation Configurations.
 SciTS Task IDASU01ASU03BIU01BIU03EAU01MEU01NEU06PHU01PHU04
Text LLMGPT-4.1-mini67.215.60.212.767.044.016.124.052.7
Gemini2.5-Flash64.116.31.512.467.660.95.820.764.8
DeepSeek-V31.112.30.05.840.259.313.628.950.7
VL LLMGPT-5-mini65.718.90.817.967.630.413.321.447.8
Gemini2.5-Flash61.615.20.98.372.564.111.622.759.0
 Intern-S1-Pro98.075.920.888.399.565.671.336.893.2
Table 3: Results on a subset of the SciTS benchmark. F1 scores are reported.
DatasetBiology-InstructionIntern-S1-Pro
DNA-cpd44.5454.60
DNA-emp8.1014.02
DNA-pd58.1882.65
DNA-tf-h24.4554.11
DNA-tf-m39.9160.80
Multi_sequence-antibody_antigen10.2644.76
Table 4: Comparison between specialized model (Biology-Instruction [ 14 ] ) and Intern-S1-Pro on biological tasks.
More in this topic

Multimodal Foundation Models

Multimodal foundation models learn shared capabilities across text, images, audio, video, and other signals for understanding, generation, and action.

See all 354 papers

Explore similar work

CardsList
Aug 13, 2026

Intern-S2-Preview: Scientific Agentic Foundation Model

Scientific discovery increasingly requires AI systems that can reason over scientific evidence of heterogeneous modalities, interact with scientific tools and environments, and sustain progress across long task horizons. We present Intern-S2-Preview, a series of scientific agentic foundation models designed to support multimodal scientific understanding, reasoning, generation, and long-horizon tasks. The training pipeline begins with scientific multimodal pre-training over rendered scientific documents, interleaved image-text data, and diverse scientific corpora. Starting from the pretrained checkpoint, we apply a unified post-training pipeline consisting of supervised fine-tuning, scalable multi-task reinforcement learning (RL), black- and white-box agentic RL, and on-policy distillation. This pipeline is supported by practical techniques that improve rollout and training stability and efficiency, including partial rollout with off-policy correction, adaptive length regularization, online speculative decoding, robust multi-task optimization, and trace-aware experience assembly for agentic tasks. At the architecture level, Intern-S2-Preview-397B extends time series modelling from efficient long-sequence understanding to numerical forecasting, while Memory Decoder is studied as a separate memory-augmented path for rapid scientific specialization without modifying the frozen 397B backbone. Evaluations across scientific, multimodal, agentic, and general-purpose benchmarks show that Intern-S2-Preview-397B achieves competitive or leading results in multiple settings. The time series modules improve scientific signal understanding and forecasting on SciTS, while the separate Intern-MemDec-4B extension improves the Biology-Instructions average score from 56.92 to 60.32 without modifying the frozen 397B backbone.

Lei Bai, Jiaqi Cao, Chiyu Chen +122

AI Agents for Scientific DiscoveryTime Series ForecastingMultimodal ReasoningMemory-Augmented Language Model Agents
AskPDF
Jul 17, 2026

S1-Omni: A Unified Multimodal Reasoning Model for Scientific Understanding, Prediction, and Generation

We present S1-Omni, a unified multimodal reasoning model for scientific understanding, prediction, and generation. AI for Science (AI4S) has advanced significantly through domain-specific models, tool-augmented LLMs, and scientific language models. However, model capabilities remain highly fragmented, limiting the joint modeling of heterogeneous data, scientific laws, and expert knowledge. S1-Omni addresses this gap by consolidating these capabilities into a single, coherent scientific reasoning model. The architecture of S1-Omni is built upon three core components: unified representation of scientific data, natural-world knowledge alignment, and decoding for domain-specific tasks. First, S1-Omni maps natural-language instructions and scientific objects, including CIF, SMILES, protein sequences, spectra, and scientific images, into a shared representation space. Second, it incorporates scientific laws and expert knowledge into data construction and training, enabling the model to reason from scientific evidence. Third, it performs task-specific decoding to support a broad range of applications, including property prediction, spectrum-to-molecular generation, protein site and structure prediction, and scientific image generation and editing. S1-Omni is trained on S1-Omni-Corpus, which covers 200 scientific tasks and contains millions of reasoning samples, and is evaluated on over 60 scientific benchmarks. It outperforms GPT-5.5 and Gemini-3.1-Pro on most benchmarks and matches or surpasses domain-specific models on several benchmarks. Overall, S1-Omni provides a practical path toward unified scientific modeling.

Jiahao Zhao, Junyi Liu, Lifeng Xu +19

ScienceOne AI

AI for ScienceUnified Multimodal ModelsMultimodal ReasoningMultimodal Generation
AskPDF
Jun 14, 2026

SciOrch: Learning to Orchestrate Expert LLMs for Solving Frontier Multimodal Scientific Reasoning Tasks

Frontier scientific reasoning remains a major challenge for large language models (LLMs), where even the strongest commercial systems fall short of expert-level performance. A closer look at model behavior reveals substantial complementarity that single-model evaluation hides: different frontier models excel on different question types, and no single model captures the full picture. We present SciOrch, a framework that trains a lightweight 8B model to orchestrate frontier LLMs for scientific reasoning. The orchestrator decomposes each question, delegates sub-problems to selected commercial models through API calls, and synthesizes a final answer. Training such an orchestrator is fundamentally harder than conventional agentic RL: each action triggers an API call that is expensive in both dollar cost and latency, making standard online rollouts infeasible. We address this with MCTS-based approach, producing diverse orchestration trajectories, extracting per-node single-turn samples, and optimizing the orchestrator with GRPO-style training. On a 240-question test set spanning SGI-Reasoning and Scientists' First Exam, SciOrch reaches 56.66% average accuracy, outperforming the strongest single commercial model by 3.74% and the strongest multi-agent baseline by 3.33%. It also attains the best accuracy on both SGI and SFE with less than half the API cost of typical multi-agent methods.

Jingru Guo, Xiangyuan Xue, Lian Zhang +6

1Imperial College London · 2The Chinese University of Hong Kong · University of Illinois Urbana-Champaign +4

LLM Agent OrchestrationMultimodal ReasoningScientific Reasoning in Language ModelsLarge Language Model-Guided Scientific DiscoveryRL for LLM Agents
AskPDF