🚀 Introducing Hunyuan-A13B, our latest open-source LLM.
As an MoE model, it leverages 80B total parameters with just 13B active, delivering powerful performance that scores on par with o1 and DeepSeek across multiple mainstream benchmarks.
Hunyuan-A13B features a hybrid architecture with dynamic 'fast and slow' reasoning, excels at long-text tasks, and has advanced agentic tool-calling capabilities for executing complex tasks with natural language.
To further improve the capabilities of LLMs, we've also open-sourced two new datasets: ArtifactsBench is designed to bridge the visual and interactive gap in code evaluations, while C3-Bench is an agent-specific evaluation dataset designed to reveal model vulnerabilities and promote research into performance interpretability.
We're open-sourcing Hunyuan-A13B to foster community-driven innovation. Explore the corresponding API and resources below:
Try it on: https://t.co/U3966JuQHi
API Address: https://t.co/ED2ZFX56v5
GitHub: https://t.co/ZkNqmeOW5m
Hugging Face: https://t.co/YJ5pXtRoZv
C3-Bench Dataset: https://t.co/uA19fGwTD8
ArtifactsBench Dataset: https://t.co/Sq63ffAn02