📂 Avatars 👁 2.8k views 🕐 May 24, 2026

Tongyi DeepResearch

Tongyi DeepResearch is the first fully open-source Web Agent to achieve performance.

Tongyi DeepResearch is the first fully open-source Web Agent to achieve performance on par with OpenAI’s DeepResearch across a comprehensive suite of benchmarks. It is designed for researchers and developers looking to create autonomous agents that can perform complex information-seeking tasks. Tongyi DeepResearch demonstrates state-of-the-art results, scoring 32.9 on the academic reasoning task Humanity’s Last Exam (HLE), 43.4 on BrowseComp, and 46.7 on BrowseComp-ZH.

Tongyi DeepResearch works by utilizing a novel data synthesis solution applied across the entire training pipeline, from Agentic Continual Pre-training (CPT) and Supervised Fine-Tuning (SFT) for cold-starting, to the final Reinforcement Learning (RL) stage. The model is trained using a customized on-policy Group Relative Policy Optimization (GRPO) algorithm, which ensures that the learning signal is always relevant to the model’s current capabilities. The training objective is optimized using a token-level policy gradient loss, and the model is evaluated on a range of benchmarks, including Humanity’s Last Exam (HLE), BrowseComp, and xbench-DeepSearch.

Tongyi DeepResearch is most valuable for researchers and developers who need to create autonomous agents that can perform complex information-seeking tasks. The model’s ability to synthesize web-based QA data through a novel pipeline and its capacity to systematically outperform existing proprietary and open-source Deep Research agents make it an ideal choice for those looking to create advanced agents. The model’s open-source nature also makes it an attractive option for those who want to customize and extend the model to suit their specific needs.

Avatars Business Ai Chat Ai
Visit Official Site Freemium
Features
◈
Continual Pre-training and Post-training Empowered by Fully Synthetic Data
allows the model to learn from synthetic data and improve its performance on complex tasks
⟐
On-Policy Agent Reinforcement Learning (RL)
enables the model to learn from its interactions with the environment and improve its decision-making capabilities
⬡
End-to-End Agent Training Pipeline
provides a comprehensive training pipeline that includes data synthesis, model training, and evaluation
◎
Real-World Applications and Impact
can be applied to a range of real-world tasks, including information retrieval, question answering, and decision-making
Verdict
Best forTeams doing Avatars work who need consistent output without a steep learning curve.
Skip ifYou only need this once or twice; the subscription cost won't pay off for occasional use.
✓Achieves state-of-the-art results on academic reasoning tasks and outperforms existing agents
✓Utilizes a novel data synthesis solution that allows for the creation of complex and realistic data
✓Is fully open-source, making it customizable and extensible
✕Requires significant computational resources and expertise to train and deploy
✕May not perform well on tasks that require common sense or real-world experience
Alternatives
ToolPricingUpvotesRating
Read AI Freemium ▲ 112 ★ 3.7
BigIdeasDB Freemium ▲ 315 ★ 3.5
Juice AI Freemium ▲ 280 ★ 4.1
Frequently Asked Questions
Tongyi DeepResearch is a fully open-source Web Agent that achieves state-of-the-art results on academic reasoning tasks and outperforms existing agents. It is designed for researchers and developers looking to create autonomous agents that can perform complex information-seeking tasks.
The benefits of using Tongyi DeepResearch include its ability to achieve state-of-the-art results on academic reasoning tasks, its novel data synthesis solution, and its fully open-source nature, making it customizable and extensible.
Tongyi DeepResearch works by utilizing a novel data synthesis solution applied across the entire training pipeline, from Agentic Continual Pre-training (CPT) and Supervised Fine-Tuning (SFT) for cold-starting, to the final Reinforcement Learning (RL) stage. The model is trained using a customized on-policy Group Relative Policy Optimization (GRPO) algorithm.
The use cases for Tongyi DeepResearch include creating autonomous agents that can perform complex information-seeking tasks, such as answering academic questions or retrieving information from the web, as well as creating chatbots or virtual assistants that can provide personalized recommendations or answer user queries.
Tongyi DeepResearch systematically outperforms existing proprietary and open-source Deep Research agents on a range of benchmarks, including Humanityu2019s Last Exam (HLE), BrowseComp, and xbench-DeepSearch.
Reviews
📝
No reviews yet
Be the first to share your experience with Tongyi DeepResearch.
Submit a Review

Your email address will not be published. Required fields are marked *

Tongyi DeepResearch
Tongyi DeepResearch
Freemium
Visit Site ↗
Home Prompts