Shanghai AI Lab releases Sage-A1.5, a 35B MoE agentic model for long-horizon tasks and scientific discovery
InternScience · hf · 2026-10-21
Shanghai Artificial Intelligence Laboratory introduces Sage-A1.5, a 35B mixture-of-experts agentic model that advances long-horizon problem solving over A1 via adaptive capability composition. Three key improvements:
- Agentic orchestration: extends single-agent execution with task delegation and parallel subagent collaboration under isolated contexts, decomposing complex tasks concurrently while reducing interference.
- Agentic verification: strengthens long-horizon reliability through progress checking, evidence validation, and reflective strategy revision, letting the agent abandon unproductive search paths.
- Open scientific discovery: moves beyond answering predefined questions to conducting exploratory research connected with interactive analysis and inspectable research artifacts.
More from Models
- One Week After Gemini 4 Argon Launch, Users Are Building 3D Worlds, Physics Sims and Games — minchoi · 2026-10-21
- Minimax H3 local workflow is '0/10': hyper-sensitive prompts and VRAM failures on 24GB — Ambitious_Corgi5723 · 2026-10-21
- ChatGPT's new UI keeps attaching wrong photos to real people, user reports — ProfessionalAd1198 · 2026-10-21
- Tone-filtering AI is "preemptive shame applied at machine speed" — PierceLilholt · 2026-10-21
- Was Opus quietly nerfed? Same prompt, 20 days apart, noticeably worse output — JiliJeanlouis · 2026-10-21
- Users hit a new Claude friction: paternalistic redaction of 'sensitive' data, no explanation — mattbeane · 2026-10-21