AI Breakthrough: Tencent Unveils Hunyuan Turbo

Tencent Launches Hunyuan Turbo S AI Model

• Hunyuan Turbo S doubles word generation speed and reduces first-word delay by 44%.
• The model integrates Mamba and Transformer technologies in a Mixture of Experts (MoE) architecture.
• Performance scores include 89.5 on MMLU, surpassing OpenAI’s GPT-4o, and top marks in MATH and AIME2024.
• Chinese language task score: 70.8 on Chinese-SimpleQA, outperforming DeepSeek’s 68.0.
• Pricing: 0.8 yuan ($0.11) per million tokens for input and 2 yuan ($0.28) for output.

Tencent’s Hunyuan Turbo S model demonstrates faster processing speeds with a hybrid architecture, achieving high performance benchmarks like an MMLU score of 89.5, while being priced competitively at $0.11 per million input tokens.