Tencent has announced that its large language model (LLM), “Hunyuan”, is now available for enterprises to use. The LLMs offered by technology giants, such as Google, Meta and OpenAI, are used by businesses to develop ChatGPT-like artificial intelligence (AI) chatbots.
The Chinese tech giant, which also owns WeChat, demoed the working of the model, and said Hunyuan had become the foundation of more than 50 of its products and services.
“By July, there are more than 130 large language models in China. A war of a hundred models has begun,” news agency Reuters quoted Jiang Jie, Tencent’s vice president, as saying.
Tencent said its model is capable of conversing in both Chinese and English and is “better” than OpenAI’s ChatGPT in areas, which include writing long text with thousands of words and solving certain math problems.
Tokens and parameters game
Tencent highlighted that its LLM has more than 100 billion parameters and was trained with more than 2 trillion tokens. These two metrics are often used to measure AI models’ power and complexity.
In comparison, OpenAI’s GPT-3 AI model contained 175 billion parameters in 2020 and Meta Platform Inc’s Llama 2 model had 70 billion parameters in 2023. Reports suggested that Google’s PaLM 2 LLM that was unveiled at Google I/O, is trained on 3.6 trillion tokens and has 340 billion parameters.
Tencent claimed that its model experiences 30% less hallucination compared to Llama 2. Hallucination is a concept where AI models generate incorrect information but present it as if it is a fact.
Recently, several Chinese tech firms, including Baidu Inc and SenseTime Group unveiled their own AI models.
The Chinese tech giant, which also owns WeChat, demoed the working of the model, and said Hunyuan had become the foundation of more than 50 of its products and services.
“By July, there are more than 130 large language models in China. A war of a hundred models has begun,” news agency Reuters quoted Jiang Jie, Tencent’s vice president, as saying.
Tencent said its model is capable of conversing in both Chinese and English and is “better” than OpenAI’s ChatGPT in areas, which include writing long text with thousands of words and solving certain math problems.
Tokens and parameters game
Tencent highlighted that its LLM has more than 100 billion parameters and was trained with more than 2 trillion tokens. These two metrics are often used to measure AI models’ power and complexity.
In comparison, OpenAI’s GPT-3 AI model contained 175 billion parameters in 2020 and Meta Platform Inc’s Llama 2 model had 70 billion parameters in 2023. Reports suggested that Google’s PaLM 2 LLM that was unveiled at Google I/O, is trained on 3.6 trillion tokens and has 340 billion parameters.
Tencent claimed that its model experiences 30% less hallucination compared to Llama 2. Hallucination is a concept where AI models generate incorrect information but present it as if it is a fact.
Recently, several Chinese tech firms, including Baidu Inc and SenseTime Group unveiled their own AI models.