# InternScience releases Sage-A1.5 35B agentic model for scientific tasks

Digest AI · Generative AI & Models · published 2026-10-21T08:59:41Z

Canonical: https://digestai.news/story/internscience-releases-sage-a1-5-35b-agentic-model-for-scientific-task

## Summary

InternScience has published Sage-A1.5, a 35 billion‑parameter Mixture‑of‑Experts model built on Shanghai AI Laboratory’s Qwen3.6‑35B‑A3B. The model is packaged for Hugging Face Transformers and can be served with vLLM or SGLang, supporting a 200 K‑token context window, tool calling, and two optional modes – Agentic Orchestration and Verification.

In the authors’ offline evaluation the model achieved top scores on several benchmarks, including 92.0 on XBench‑DS‑2510, 81.6 on IFBench, 90.0 on PinchBench, 81.0 on FrontierScience‑Olympiad, 48.3 on FrontierScience‑Research, 80.0 on CMT, 91.5 on OmniDocBench and a GAIA score of 95.1 that exceeds every larger model tested. Compared with its predecessor A1, Sage‑A1.5 improves SWE‑bench Pro by +17.6 points and several other agentic and engineering metrics.

The repository provides weights, configuration files, and example launch commands for both standard and speculative multi‑token prediction serving. Users can enable the orchestration mode by registering a `spawnsubagent` tool, and turn on verification via a chat‑template flag. The release is intended to give the community an open foundation for long‑horizon scientific problem solving.

## Key points

- Sage-A1.5 is a 35B Mixture‑of‑Experts agentic model based on Qwen3.6‑35B‑A3B, released with Hugging Face weights.
- It scores 92.0 on XBench‑DS‑2510 and 95.1 on GAIA, outperforming larger frontier models on several benchmarks.
- The repo includes serving scripts for vLLM and SGLang, a 200K‑token context window, and optional orchestration and verification modes.

## Why it matters

An open‑source 35B agentic model that rivals larger systems lowers the barrier for building advanced scientific AI agents.

## Sources

1. [InternScience/Sage-A1.5 Finetune of Qwen 3.6 35b-a3b](https://huggingface.co/InternScience/Sage-A1.5) (huggingface.co, 2026-10-21, primary source)

Part of the developing story: [AI Agents Rise Across Finance and Science](https://digestai.news/thread/navalia-improves-cowork-agent-performance-over-qwen3-6-35b-a3b-on-finance) (2 stories)

## Cite

Digest AI, "InternScience releases Sage-A1.5 35B agentic model for scientific tasks", 21 October 2026, https://digestai.news/story/internscience-releases-sage-a1-5-35b-agentic-model-for-scientific-task

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/internscience-releases-sage-a1-5-35b-agentic-model-for-scientific-task.json