---
title: AgenticQwen-30B-A3B
canonical_url: "https://www.modelscope.ai/models/PAI/AgenticQwen-30B-A3B"
md_url: "https://www.modelscope.ai/models/PAI/AgenticQwen-30B-A3B.md"
repository: PAI/AgenticQwen-30B-A3B
last_updated: 2026-03-17
model_type:
  - qwen3_moe
architectures:
  - Qwen3MoeForCausalLM
parameters: 30.5B
tensor_type:
  - BF16
library_name:
  - safetensors
  - pytorch
frameworks:
  - Pytorch
downloads: 168
stars: 0
---

# AgenticQwen-30B-A3B

> AgenticQwen-30B-A3B - An open-source model by PAI on ModelScope. AgenticQwen-30B-A3B is a small agentic language model trained on Qwen3-30B-A3B-Instruct, designed for multi-step reasoning and tool use. It is trained with a multi-round reinforcement learning…

- **Repository**: PAI/AgenticQwen-30B-A3B
- **Parameters**: 30.5B
- **Downloads**: 168
- **Stars**: 0
- **Last updated**: 2026-03-17

Source: https://www.modelscope.ai/models/PAI/AgenticQwen-30B-A3B

---

# Model Description

AgenticQwen-30B-A3B is a small agentic language model trained on Qwen3-30B-A3B-Instruct, designed for multi-step reasoning and tool use. It is trained with a multi-round reinforcement learning (GRPO-style) pipeline and a dual "data flywheel" mechanism that continually increases task difficulty for both reasoning and agentic workflows.