InternScience/Sage-A1.5
AI model · https://huggingface.co/InternScience/Sage-A1.5
First seen by Oriel on .
AI summary
Sage-A1.5 is a 35B Mixture-of-Experts agentic model for long-horizon scientific and professional tasks, with optional parallel subagents and in-turn self-review.
- The model supports tool use, task decomposition, parallel subagents with isolated contexts, and strategy revision based on progress checks and evidence validation.
- It natively supports a 256K context window; the page documents YaRN configurations for serving at 512K with SGLang or vLLM.
- The authors report scores including 95.1 on GAIA, 61.2 on SWE-bench Pro, and 48.3 on FrontierScience-Research, alongside comparisons with larger models.
- Agentic Orchestration is enabled through a registered spawn_subagent tool, while Verification Mode is enabled per request with chat_template_kwargs and is off by default.
#llm agents #mixture of experts #scientific reasoning #vllm
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Likes
Total likes reported by the source.
- Hugging Face Models: 379, latest observation ; +266 over observed period since
Downloads (30d)
Rolling 30-day downloads. A change in this value is not new downloads during the observed period.
- Hugging Face Models: 11,029, latest observation ; +10,971 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Models | InternScience/Sage-A1.5 | https://huggingface.co/InternScience/Sage-A1.5 Read discussion |