On March 21, Tencent once again dropped a "bombshell" in the field of AI. Following the launch of a new generation of fast-thinking model hybrid element Turbo at the end of February After S, Tencent announced the launch of the official version of the self-developed deep thinking model hybrid element T1, which was simultaneously launched on the official website of Tencent Cloud. This news quickly attracted widespread attention and heated discussions in the industry.
|Mixed element T1 scores are industry-leading
The release of the official version of Tencent Mixed Yuan T1 can be said to be "as scheduled".
Previously, in the hybrid element Turbo When S was officially released, Tencent revealed that the official version of Tencent's hybrid T1 model API will also be launched soon to provide access services to the outside world. Now, that promise has finally been fulfilled.
Tencent mixed element summary,Mixed elementT1The official version is characterized by"fast pronunciation、Can reply in seconds","Good at ultra-long text processing",It is a strong inference model developed by Tencent。
In terms of speed, the decoding performance of the official version of the hybrid element T1 is 2 times higher than that of the industry under comparable parameters, the first word is out in seconds, and the speed of outgoing words is 60 to 80 tokens per second, with excellent long text processing capabilities. A number of indicators have reached the level of industry-leading inference models, and in terms of effect, the official version of Hybrid T1 has strong logic, concise style, accurate follow-up of complex instructions, and low abstract hallucination.
Through large-scale reinforcement learning, combined with special optimization of science problems such as mathematics, logical reasoning, science, and code, the official version of Hybrid T1 further improves the reasoning ability.
Among the common evaluation indicators that reflect the basic capabilities of inference models, such as the large language model (LLM) evaluation enhancement dataset MMLU-PRO, Hybrid T1 scored 87.2, second only to OpenAI with a score of 89.3 o1, which is higher than OpenAI, which scored 86.1 GPT 4.5 and DeepSeek with a score of 84 R1。
Tencent's hybrid model continues to iterate rapidly
At present, Mixed Yuan T1 has been launched on the official website of Tencent Cloud, and in terms of pricing, the input price is per million tokens 1 yuan, and the output price is per million tokens $4. This price strategy will undoubtedly attract more developers and enterprise users to try and use the hybrid T1, and further promote its wide application in the AI field.
Since the beginning of this year, Tencent's hybrid model has continued to iterate rapidly, and AI seems to have become the biggest driving force within Tencent. Tencent has successively launched the deep thinking model T1 Preview and Fast Thinking Model Turbo S, and has been widely used in Tencent ingots, ima, Tencent Docs, WeChat reading, Sogou input method, QQ browser and other internal products. This move not only improves the intelligence level of Tencent products, but also brings users a more convenient and efficient user experience.
At the same time, Tencent Hybrid actively embraces open source, and its open source model comprehensively covers multiple modalities such as text, image, video, and 3D generation. Behind this strategic choice, on the one hand, Tencent has long adhered to the concept of technology inclusion, hoping to benefit more developers and users with the achievements of AI technology. On the other hand, it also benefits from Tencent's multi-scenario and ecological advantages in users and industries, providing a broad space for the implementation of AI technology.
In the 2024 financial report released a few days ago, Tencent's annual R&D investment of 70.7 billion yuan set a record. Among them, the boosting effect of AI technology on WeChat's sprout business is particularly significant, promoting the rapid growth of Tencent's marketing service segment. In the fourth quarter of 2024, Tencent's capital expenditure in the field of AI exceeded 39 billion yuan, and it announced that it will continue to increase its weight in 2025, and its AI strategy has entered a period of heavy investment.
Major Internet giants have increased their investment and layout in the field of AI
2025 can be said to be a big year for China's technological breakthroughs, and major Internet giants have increased their investment and layout in the field of AI, and the competition is extremely fierce. Recently, ByteDance's Doubao large model team announced the open-sourcing of a key optimization technology for the MoE (Hybrid Expert Model) architecture, which can improve the training efficiency of large models by 1.7 times while saving 40% of the cost. The launch of this technology has undoubtedly brought ByteDance's competitiveness in the field of AI to a higher level.
Alibaba has not been idle, and recently launched its flagship AI application, New Quark, which focuses on "deep thinking" capabilities. After the user enters the command in the "AI super box", the Quark intelligent center can automatically identify the intent, plan the task, and mobilize different models and agent modules to help the user complete the task. This innovative application brings users a smarter and more convenient search experience, which can be said to be an important implementation of AI technology.
Baidu is also making frequent moves, and has released two free new models one after another - Wenxin Model 4.5 and Wenxin Model X1. Wenxin Model 4.5 has significantly improved its multimodal understanding ability, while Wenxin Model X1 is a deep thinking model, with performance benchmarking DeepSeek-R1, with "long thinking chain" ability, good at Chinese knowledge question answering, literary creation and logical reasoning and other tasks, while also greatly reducing the cost of reasoning.
In general, 2025 is the year of the explosion of AI technology in China, and giants such as ByteDance, Alibaba, and Baidu have shown their "killer features" to promote the innovation and implementation of AI technology. The popularization of open source technology and the deepening of commercial applications will bring more opportunities to the AI industry chain.
Investors can focus on the application of AI technology in the Internet, search, content creation, and other fields, after all, the wave of AI has only just begun.