1d:I[70616,["/_next/static/chunks/0kcf5dt6yxdox.js","/_next/static/chunks/1su5n1uv1ti4s.js","/_next/static/chunks/0gvke1wv6udfg.js","/_next/static/chunks/1dfrle3jgmlzh.js","/_next/static/chunks/162lgvduyr-fa.js","/_next/static/chunks/2gcnnopam8e9w.js","/_next/static/chunks/2g60iab5a1pkg.js"],"Breadcrumb"] 1e:I[90762,["/_next/static/chunks/0kcf5dt6yxdox.js","/_next/static/chunks/1su5n1uv1ti4s.js","/_next/static/chunks/0gvke1wv6udfg.js","/_next/static/chunks/1dfrle3jgmlzh.js","/_next/static/chunks/162lgvduyr-fa.js","/_next/static/chunks/2gcnnopam8e9w.js","/_next/static/chunks/2g60iab5a1pkg.js"],"ArticleGrid"] 1f:T1055,

Researchers behind Shanghai AI Laboratory's Intern model family have published Intern-S2-Preview, a series of foundation models built specifically for scientific work that requires reasoning over evidence in several modalities, operating scientific tools and environments, and sustaining progress across long task horizons.

The flagship, Intern-S2-Preview-397B, carries 397 billion parameters. The paper, posted to arXiv on August 13, lists more than 120 authors — an indication of the scale of the engineering effort behind it.

A pipeline built for long horizons

Training begins with scientific multimodal pretraining over rendered scientific documents, interleaved image-text data and broad scientific corpora. From that pretrained checkpoint the team applies a unified post-training pipeline: supervised fine-tuning, scalable multi-task reinforcement learning, both black-box and white-box agentic RL, and on-policy distillation.

Much of the paper's contribution lies in making that pipeline stable at scale. The authors describe partial rollout with off-policy correction, adaptive length regularization, online speculative decoding, robust multi-task optimization and trace-aware experience assembly for agentic tasks — a set of techniques aimed squarely at the rollout and training instabilities that make long-horizon reinforcement learning expensive.

Two architectural bets

Intern-S2-Preview-397B extends time series modelling beyond efficient long-sequence understanding into numerical forecasting, which matters for domains where the raw evidence is a signal rather than a document or an image. The team reports that these modules improve scientific signal understanding and forecasting on the SciTS benchmark.

The second bet is Memory Decoder, studied as a separate memory-augmented path rather than a change to the main model. The goal is rapid scientific specialization without touching the frozen 397B backbone. The team reports that a separate four-billion-parameter extension, Intern-MemDec-4B, lifted the average Biology-Instructions score from 56.92 to 60.32 — a gain achieved without modifying the underlying model at all.

That decoupling is arguably the most practically interesting result in the paper. Specializing a frontier-scale model for a narrow scientific domain normally means fine-tuning it, which is costly and produces a separate copy for every field. A small attachable module that delivers a comparable lift leaves the base model untouched and shareable across domains.

Where it lands

Across scientific, multimodal, agentic and general-purpose benchmarks, the authors report competitive or leading results in multiple settings — careful phrasing that stops short of claiming a clean sweep. The model is labelled a preview, and the paper reads as a progress report on an architecture still being explored rather than a finished product.

The framing is the notable part. Most frontier model releases are judged on coding, general reasoning and consumer-facing chat quality. Intern-S2-Preview aims at a narrower target: systems that read scientific evidence in several modalities at once, call scientific tools, and keep working through tasks that unfold over many steps rather than a single exchange.

That target has become a crowded ambition. Laboratories across the industry have been positioning agentic models as instruments for research itself rather than assistants to researchers. What separates the proposals is less the goal than the mechanics — how training stays stable across long agentic rollouts, and how a general model gets specialized without being rebuilt. On both counts, Intern-S2-Preview offers a concrete account.

The abstract does not specify release timing or licensing terms for the model weights, and the preview label suggests the series is still evolving. For researchers weighing whether to look closer, the numbers that carry the most information are the SciTS forecasting gains and the Biology-Instructions jump, because both test the architecture's specific claims rather than general capability.

a:["$","div",null,{"className":"mx-auto max-w-7xl px-4 py-8 sm:px-6","children":[["$","script",null,{"type":"application/ld+json","dangerouslySetInnerHTML":{"__html":"{\"@context\":\"https://schema.org\",\"@type\":\"BreadcrumbList\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https://ainewsway.com\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"#shanghai-ai-laboratory\",\"item\":\"https://ainewsway.com/tag/shanghai-ai-laboratory\"}]}"}}],["$","script",null,{"type":"application/ld+json","dangerouslySetInnerHTML":{"__html":"{\"@context\":\"https://schema.org\",\"@type\":\"ItemList\",\"@id\":\"https://ainewsway.com/tag/shanghai-ai-laboratory#itemlist\",\"name\":\"#shanghai-ai-laboratory Articles\",\"numberOfItems\":1,\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"url\":\"https://ainewsway.com/articles/intern-s2-preview-puts-397-billion-parameters-behind-scientific-agents-mssq9arx\",\"name\":\"Intern-S2-Preview Puts 397 Billion Parameters Behind Scientific Agents\"}]}"}}],["$","$L1d",null,{"items":[{"label":"# shanghai-ai-laboratory"}]}],["$","h1",null,{"className":"mb-8 text-3xl font-bold text-zinc-900 dark:text-zinc-100","children":["#","shanghai-ai-laboratory"]}],[["$","$L1e",null,{"articles":[{"id":"b510c7c4-c83e-4926-af8e-cec5365e6857","slug":"intern-s2-preview-puts-397-billion-parameters-behind-scientific-agents-mssq9arx","title":"Intern-S2-Preview Puts 397 Billion Parameters Behind Scientific Agents","subtitle":"The Intern team pairs multimodal scientific pretraining with agentic reinforcement learning, and adds a memory module that specializes the model without retraining it","content":"$1f","excerpt":"Intern-S2-Preview-397B pairs multimodal scientific pretraining with agentic RL, and a separate memory module lifts Biology-Instructions from 56.92 to 60.32.","coverImage":"https://cdn.ainewsway.com/processed/v1/1f916836-56f6-4b30-9255-17b859906a24/cover.webp","coverImageCaption":"Racks inside the Frontier supercomputer at Oak Ridge National Laboratory, illustrating the class of compute infrastructure that training a 397-billion-parameter scientific model requires.","authorId":"6773f67d-d7fa-4b19-8f38-384f10ac761b","status":"PUBLISHED","publishedAt":"2026-08-14T09:11:12.996Z","viewCount":30,"likeCount":0,"readingTime":3,"isFeatured":false,"isBreakingNews":false,"metaTitle":"Intern-S2-Preview 397B: Shanghai AI Lab's Scientific Model","metaDescription":"Shanghai AI Lab's Intern-S2-Preview-397B pairs scientific multimodal pretraining with agentic RL; a 4B memory decoder lifts Biology-Instructions to 60.32.","metaKeywords":["Intern-S2-Preview","Shanghai AI Laboratory","scientific AI model","agentic reinforcement learning","Memory Decoder","foundation models","ai-ml"],"ogImage":"https://cdn.ainewsway.com/processed/v1/1f916836-56f6-4b30-9255-17b859906a24/og.jpg","aiSummary":"Shanghai AI Laboratory researchers published Intern-S2-Preview on arXiv on August 13, a family of scientific foundation models led by a 397-billion-parameter flagship trained with multimodal scientific pretraining and agentic reinforcement learning. A separate 4-billion-parameter Memory Decoder module, Intern-MemDec-4B, raised the Biology-Instructions average from 56.92 to 60.32 without changing the frozen backbone. The decoupled specialization approach could replace costly per-domain fine-tuning.","aiTags":[],"sentiment":null,"articleType":"NEWS","videoEmbed":null,"videoType":null,"seoScore":85,"isAutoGenerated":true,"isBackfill":false,"pipelineStage":null,"copyrightScore":null,"similarityScore":null,"matchedSourceId":null,"createdAt":"2026-08-14T09:11:12.334Z","updatedAt":"2026-09-10T01:38:00.887Z","contentUpdatedAt":"2026-08-14T09:11:12.996Z","deletedAt":null,"author":{"id":"6773f67d-d7fa-4b19-8f38-384f10ac761b","name":"Seung Jung","username":"Seung Jung","avatar":null},"categories":[{"id":"e2b5fffc-de32-42aa-86fe-c75f219d6b4b","name":"AI & Machine Learning","slug":"ai-ml"}],"tags":[{"id":"23d63a7b-896d-48de-a4ff-870d32e10654","name":"Intern-S2","slug":"intern-s2"},{"id":"3ab99e79-8ded-4954-b8b8-482f1c6cbbfa","name":"Shanghai AI Laboratory","slug":"shanghai-ai-laboratory"},{"id":"2156a72f-a2f0-4cb8-8c52-299093c2f33b","name":"Foundation Models","slug":"foundation-models"},{"id":"4f956424-4af0-45c1-b2e0-0afa4308426f","name":"Scientific AI","slug":"scientific-ai"},{"id":"fad4822e-d84d-42e7-9deb-bbec0f3723e2","name":"Agentic AI","slug":"agentic-ai"}],"translations":[],"celebrities":[{"id":"e6e59771-9b65-46cf-adcc-30805293306c","nameKo":"상하이 AI 연구소","nameEn":"Shanghai AI Laboratory","slug":"shanghai-ai-laboratory","profileImage":null}]}]}],"$L20"]]}]

#shanghai-ai-laboratory

Intern-S2-Preview Puts 397 Billion Parameters Behind Scientific Agents

Subscribe to Newsletter

#shanghai-ai-laboratory

Intern-S2-Preview Puts 397 Billion Parameters Behind Scientific Agents