Description
Agents-A1 is a 35B Mixture-of-Experts agentic model developed by InternScience. It is designed to scale heterogeneous agentic abilities across various domains, including long-horizon search, engineering, scientific research, instruction following, and tool-calling. The model is trained using a domain-grounded knowledge-action infrastructure that constructs actions, observations, and verifier outcomes, turning the agent's process into a trainable target.
Read More
The training paradigm for Agents-A1 involves three stages. Initially, full-domain supervised fine-tuning aligns the base model with broad agentic behaviors. This is followed by training domain-level teacher models to capture specialized expertise in each domain. Finally, multi-teacher multi-domain on-policy distillation with heterogeneity-aware optimization is employed to enhance knowledge transfer efficiency across different domains.
Agents-A1 supports function calling and tool integration, enabling seamless interaction with APIs, code interpreters, search engines, and other external tools. It handles tool-integrated scientific reasoning and professional knowledge question answering, and it follows detailed, multi-constraint instructions across diverse domains.
The model has been evaluated in real-world agentic and research-oriented workflows across six directions, delivering highly competitive performance against frontier-scale systems such as GPT-5.5 and DeepSeek-V4-pro. It achieves state-of-the-art results on several challenging benchmarks, including Seal-0, HiPhO, FrontierScience-Olympiad, and IFEval, among others. These results demonstrate Agents-A1's strong long-horizon search ability, robust scientific reasoning, and reliable instruction following, establishing it as a highly capable and efficient agentic model.
- CategoryAI Models & LLMs
- Pricing ModelFree
- DeploymentCloud
- PlatformsWeb
Agents-A1 Highlights
35B Mixture-of-Experts model
Long-horizon search capability
Tool integration support
Scientific reasoning
Instruction following
Domain-grounded knowledge-action infrastructure
Multi-teacher multi-domain distillation
State-of-the-art benchmark performance
Getting Started with Agents-A1
Access model page: Visit the Hugging Face repository for Agents-A1
Authenticate: Ensure you have the necessary credentials to access the model
Configure: Set up the model parameters and environment
Send prompts: Interact with the model using your desired inputs
Fine-tune: Adjust the model for specific tasks if needed
Agents-A1's Use Cases
- Long-horizon Search
- Engineering Tasks
- Scientific Research
- Instruction Following
- Tool Integration
