AI4SGI/ExoMind-9B

VISIONPricing:Input $0.431 / Cached $0.0862 / Output $1.12Concurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 13, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

ExoMind-9B is a 9 billion parameter language model developed by AI4SGI and the ExoMind Team at Shanghai Artificial Intelligence Laboratory, fine-tuned from Qwen3.5-9B with a 32,768 token context length. It is designed for scientific reasoning and agentic research, focusing on workflows like source discovery, evidence grounding, and executable verification. This compact checkpoint supports resource-conscious experimentation within an extended-mind-inspired agentic system, retaining multimodal capabilities.

Loading preview...

ExoMind-9B: A Compact Model for Scientific Agentic Research

ExoMind-9B, developed by AI4SGI and the ExoMind Team at Shanghai Artificial Intelligence Laboratory, is a 9 billion parameter model fine-tuned from Qwen3.5-9B. It is specifically designed for scientific reasoning and agentic research, offering a compact checkpoint for lower-resource experimentation. The model follows an extended-mind-inspired approach, integrating the language model with specialized interaction objects and autonomous interaction processes.

Key Capabilities

  • Scientific Interaction: Facilitates workflows such as source discovery, evidence grounding, executable verification, and observation integration.
  • Progressive CoI Training: Develops intrinsic reasoning and interaction behaviors through selected pure-reasoning and interaction trajectories.
  • Multimodal Foundation: Retains the image-text capabilities inherited from its Qwen3.5 base model.
  • Compact Checkpoint: Optimized for resource-conscious experiments in scientific reasoning and agentic research.

Good For

ExoMind-9B is intended for applications requiring scientific question answering, mathematical and computational reasoning, and tool-use experiments. It is also suitable for code-assisted verification and developing resource-conscious agentic prototypes. While the larger ExoMind 35B-A3B system shows strong performance on scientific benchmarks like FrontierScience-Research and AMO-Bench, ExoMind-9B provides a more accessible option for exploring these advanced agentic paradigms.

Finetunes
5 models