Tencent’s AI chatbot Yuanbao now includes two new models: the deep reasoning model Hunyuan T1979 ArchivesDeepSeek V3-0324, the company announced early Monday. The update enhances Yuanbao’s performance in code generation, mathematical reasoning, and long-text processing. Hunyuan T1, scoring 87.2 on the Massive Multitask Language Understanding (MMLU) benchmark, officially launched on Saturday. It surpasses DeepSeek-R1’s score of 84 but falls short of OpenAI’s Qwen-1’s 89.3. Tencent has priced T1 competitively, aligning with DeepSeek’s approach. The model charges RMB 1 ($0.14) per million input tokens and RMB 4 ($ 0.55) per million output tokens. This matches DeepSeek-R1’s daytime input rate, though R1 offers a lower nighttime rate of RMB 0.25 ($0.034) per million tokens. [iThome, in Chinese, SCMP]
