Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology

Hunyuan-A13B-Instruct

Model basics

Reasoning traces

Not supported

Thinking modes

Thinking modes not supported

Context length

256K tokens

Max output length

32000 tokens

Model type

基础大模型

Release date

2025-06-27

Model file size

160.8 GB

MoE architecture

Total params / Active params

800.0B / N/A

Knowledge cutoff

No data

Hunyuan-A13B-Instruct

Open source & experience

Code license

TENCENT HUNYUAN COMMUNITY LICENSE AGREEMENT

Weights license

TENCENT HUNYUAN COMMUNITY LICENSE AGREEMENT- 免费商用授权

GitHub repo

https://github.com/Tencent-Hunyuan/Hunyuan-A13B

Hugging Face

https://huggingface.co/tencent/Hunyuan-A13B-Instruct

Live demo

https://hunyuan.tencent.com/?model=hunyuan-a13b

Hunyuan-A13B-Instruct

Official resources

Paper

Hunyuan-A13B

DataLearnerAI blog

腾讯开源Hunyuan-A13B大模型：MoE架构，混合推理（支持直接回复和带推理过程后回复），原WizardLM团队打造，评测结果超Qwen2.5-72B和Qwen3-A22B

Hunyuan-A13B-Instruct

API details

API speed

3/5

No public API pricing yet.

Hunyuan-A13B-Instruct

Benchmark Results

Hunyuan-A13B-Instruct currently shows benchmark results led by MATH (3 / 41, score 94.30), AIME 2024 (17 / 62, score 87.30), MMLU (19 / 62, score 88.17). This page also consolidates core specs, context limits, and API pricing so you can evaluate the model from benchmark results and deployment constraints together.

Thinking

All modes Normal