Alibaba Cloud's Model Update and Price Reduction
TapTechNews May 21st news, ByteDance officially released the Doubao large model on May 15th and announced that the price of the large model has entered the cent era, claiming to be 99.3% cheaper than the industry.
Today, Alibaba Cloud followed closely and dropped a bombshell: The main model Qwen-Long of Tongyi Qianwen GPT-4 level announced a price reduction. The API input price dropped from $0.02 per thousand tokens to $0.0005 per thousand tokens, a reduction of 97%. (1 yuan can buy 2 million tokens, which is equivalent to the text volume of about 5 Xinhua Dictionaries. This model supports a maximum long text input of 10 million tokens, and after the price reduction, it is about 1/400 of the GPT-4 price.)
According to TapTechNews' previous report, Alibaba Cloud officially launched the brand-new Tongyi Qianwen 2.5 large model on May 9th this year, claiming that many capabilities surpassed GPT-4.
According to the official introduction of Alibaba Cloud, Tongyi large model has served more than 90,000 enterprises through Alibaba Cloud, and the cumulative download volume of Tongyi open source model has exceeded 7 million. The landing application process of Tongyi has been accelerated and has now entered the fields of PC, mobile phones, and automobiles.
At the event site, Alibaba Cloud officially released Tongyi Qianwen 2.5, claiming capability upgrade, comprehensively surpassing GPT-4. In the Chinese context, it surpassed GPT-4 in many capabilities such as text understanding, text generation, knowledge Q&A & life advice, chatting & conversation, and security risks.
In addition, compared with Tongyi Qianwen 2.1, Tongyi Qianwen 2.5 has many capability improvements, with a 9% improvement in understanding ability, a 16% improvement in logical reasoning, a 19% improvement in instruction following, and a 10% improvement in code ability.
At the same time, the latest open source 110 billion parameter model of Tongyi Qianwen has achieved the best results in multiple benchmark evaluations, surpassing Meta's Llama-3-70B and announcing to be the most powerful model in the open source field.
