DeepYardDeepYard
A

Agents-A1

Open-source multimodal agent model with image-text reasoning on Qwen 3.5 MoE architecture

Open SourceFree

About

Agents-A1 is a multimodal agent model from InternScience built on the Qwen 3.5 Mixture-of-Experts (MoE) architecture. It processes both images and text to generate text responses, specifically optimized for agent tasks like tool use and multi-step reasoning. Includes evaluation benchmarks for measuring agent performance across various tasks, making it useful for researchers and developers building vision-enabled AI agents.

Details

Typecoding-agent
Deploymentself-hosted
Supported Models

Tags

autonomousopen-sourcemulti-agenttool-useevaluationpython