home page  >  hotspot  >  key word  >  Latest information of byte skipping  >  text

The price list of the big model of byte beating beanbag was announced: prepaid/postpaid mode was adopted from 25 yuan

2024-05-21 10:40 · Source: Home of webmaster

Message from webmaster's home (ChinaZ. com) on May 21: Byte bounce Subordinate Volcanic engine The official website recently Big model of bean curd The pricing of the model has been updated. The selling prices of different versions and specifications of the model are listed in detail. The starting price is only 25 yuan. This update comprehensively demonstrated the advantages of the general model of bean curd in terms of cost performance, especially its main model, pro-32k. Compared with other models in the industry, the price dropped by an astonishing 99%. At the same time, the TPM (the number of Tokens processed per minute) reached 2.7 to 8 times that of models of the same specification, demonstrating its outstanding Performance value ratio 。

99% lower than the industry! Price list of big model of byte beating beanbag announced: from 25 yuan

Doubao Large Model, formerly known as“ lark ”, as one of the first large models to be filed through algorithms in China, has become the domestic usage maximum One of the most widely used large-scale models. The model can process up to 120 billion Tokens of text every day, and can generate 30 million pictures, proving its powerful processing ability.

The Doubao big model family includes a variety of members, such as Pro, Lite, role playing model, speech synthesis model, voice reproduction model, speech recognition model, Wensheng diagram model and Function Call model, to meet the needs of different users. In terms of billing mode, the general model of bean curd adopts a flexible post payment mode, which calculates fees according to the actual number of Tokens used. For models with 32k and below window versions, the current limiting standards are 10K RPM (requests per minute) and 800K TPM. This standard is far superior to other mainstream models in China, ensuring that the business needs of enterprise customers are met.

99% lower than the industry! Price list of big model of byte beating beanbag announced: from 25 yuan

For the long text model, the pro and lite128k versions of the general model of Doubao also perform well. The current limiting of its model is 1K RPM and 400K TPM, greatly surpassing other models of the same specification in China. In addition, the Volcano Engine also provides a prepaid model. Enterprises can purchase TPM quotas for specific models without paying for the consumption of Tokens, so that they can plan their computing power in advance for the traffic fluctuations that enterprises may face.

Volcano Engine emphasized that the goal of the big model of Doubao is to provide customers with flexible and economical payment options, meet the needs of different enterprises through the two billing methods of post payment and pre payment, help enterprises use the big model at a very low cost, and promote the wide application and implementation of the big model.

（ report ）

  • Related recommendations
key word:

  • Recommendation Byte Doubao has its own IP universe

    As a technology content creator, my daily task is to track the latest technology and business progress. In my opinion, Doubao APP is more of a player from a giant on the card table - with a strong technical background and promotion resources regardless of cost, it is a window to understand the latest progress of byte models, and an AI product that has been carefully polished. Until the melody of "Deng Chao, gogogo" swept my information flow. In March this year, in a live broadcast slice, the blogger "Clichy" pointlessly changed his name to Doubao

    Doubao APP Technical progress AI products

  • The woman paid 2000 yuan by mistake when buying vegetables. My uncle said that he only saw 5 yuan. Later, he called the police to get it back

    On July 9, an interesting incident occurred in Sichuan due to negligence in payment. When shopping for vegetables at a roadside stall, an elder sister took out a wad of cash to pay, but she was "confused" during the payment process. She held a 5 yuan note in her left hand and a stack of cash totaling more than 2000 yuan in her right hand. However, she handed over more than 2000 yuan to the vegetable seller without realizing it. Later, when the elder sister was ready to continue to select other vegetables and pay again, she found that there was little money left. In the face of this sudden

  • Top 10 AI models: Who will dominate the rise and fall of the 2025 intelligent era?

    This paper introduces the top ten AI models and their characteristics that are currently the most attractive in the world. OpenAI's GPT-4 series is good at natural language understanding and multimodal processing; Anthropic's Claude3 series is good at long text analysis and professional writing; Google's Gemini 1.5 has powerful multimodal processing capabilities; Meta's Llama3 is an open source model with strong performance; XAI's Grok series is good at real-time information processing; Mistral AI is famous for its efficient mixed expert architecture; Cohere focuses on enterprise applications; Inflection Pi focuses on emotional companionship; China's DeepSeek and Tongyi Qianwen are prominent in the localization of Chinese texts. The article suggests selecting models according to specific needs, and recommends using professional platforms for model comparison. In the future, AI will develop in the direction of longer context and more natural multimodal interaction.

    Artificial intelligence AI large model Technological innovation

  • From 2549.15 yuan! Understanding the picture of Glory Tablet MagicPad3

    The Glory Tablet MagicPad3 is officially released tonight, with the official price starting from 2999 yuan, and all systems support the national subsidy, with the purchase price starting from 2549.15 yuan. The front is equipped with a 13.3-inch screen, 165Hz ultra-high refresh rate, eight glory oasis eye protection technologies, AI super image quality engine, and 3.2K high resolution. The body is extremely light and thin, ranging from 595g to 5.79mm. It is equipped with Qualcomm Snapdragon 8 Gen3 flagship chip, self developed dual engine antenna, and supports two-way Wi Fi 7. Built in 12450mAh Qinghai Lake battery, equipped with HONOR E2 self-developed energy efficiency enhancement chip and Glory Dujiangyan AI battery

    Glory Tablet MagicPad3 High refresh rate screen Snapdragon 8

  • 2025 latest global AI model ranking, dynamic shuffle of domestic and foreign models (recommended by real-time update platform)

    In 2025, the global AI big model competition will enter a white hot stage, and international giants such as OpenAI, Google and Chinese enterprises will face each other fiercely. The technical iteration was far more than expected, and the ranking was shuffled several times in just half a year. The current top 10 comprehensive ranking of global AI models shows that: 1) GPT-4.5 leads the way with 80.4 comprehensive points; 2) Claude3.7 takes the lead in the field of programming; 3) Gemini2.0 multimodal benchmark; 4) The reasoning speed of domestic DeepSeek R1 is increased by 3 times; 5) Ali Qwen 2.5 mathematical programming single item first. The Chinese model achieves curve overtaking through open source strategy, vertical optimization and cost revolution. For example, DeepSeek R1 only costs $6 million to reach the GPT-4 level, and the cost of byte bean packs using sparse MoE architecture is significantly reduced. The depth optimization of Chinese scenes is outstanding, for example, the accuracy of Wenxin Yiyan 4.0 dialect interaction is 92%. Open source ecology has exploded, and Tongyi Qianwen's full-size open source (7B~110B parameters) ranks first in Hugging Face. The authoritative evaluation needs to combine standardized tests, human blind tests and scene adaptation.

    Global AI big model ​OpenAI Technical iteration

  • Accuracy 92%! Apple's new model can predict pregnancy through behavior patterns

    According to media reports, Apple, together with the American Heart Association and the Brigham and Women's Hospital of Harvard Medical School, has made a major breakthrough in the field of health prediction and launched the wearable behavior model (WBM) based on behavioral data. This research marks the expansion of health monitoring from traditional biological indicators (such as heart rate and blood oxygen) to behavioral data analysis, providing a new perspective for disease prediction. The research team uses more than 2.5 billion hours of wearable device data to train the WBM model, so that it can perform high-level behaviors such as steps and activity ability

  • A roast duck sells for only 25 yuan, but the shopkeeper still makes money: the reason behind the discovery

    Recently, the topic of "one live duck costs more than one hundred yuan, and one roast duck costs only 25 yuan" hit the hot search, causing widespread public concern and discussion about the price of roast duck and the profitability of the store. After investigation, it was found that there were many ways behind the 25 yuan roast duck on the street, and the shopkeeper could still make a profit by virtue of low cost and high sales volume. Street roast duck shops mostly use Cherry Valley white feather meat duck, which has the characteristics of "eating less and growing faster", and the breeding cost is relatively low. In addition, the merchants adopt the split sales model, which is used to make the duck for roast duck

    Roast duck price Store profits Low cost and high sales volume

  • Double first! Baidu Intelligent Cloud leads the bid winning market of the big model in the first half of 2025

    In the first half of 2025, the global AI industry will usher in a key turning point, and major breakthroughs will be made in core capabilities such as multimodal understanding and in-depth logical reasoning. The big model accelerates the penetration of core industrial scenarios, and promotes the transformation of AI from general capabilities to in-depth industrial applications. China's large model projects have shown explosive growth, with 1810 winning projects, and the amount of money exceeded 6.4 billion yuan. Baidu Smart Cloud has shown outstanding performance, ranking "Double First" with 48 winning projects and 510 million yuan, and continues to lead in key industries such as finance, energy and government affairs. The industry trend shows that the big model market is moving from "general competition" to "deep industry cultivation", and technical strength and landing ability become the key. Baidu Smart Cloud, relying on leading big model technology and full stack intelligent infrastructure, helps enterprises deploy AI efficiently. Currently, 65% of central enterprises have chosen to cooperate with it. In energy, transportation, finance and other fields, Baidu has created a number of industry benchmark cases to significantly improve business efficiency.

    AI industry Multimodal understanding Large model application

  • OpenAI core model adopts physical isolation: explicit permission is required to prevent secret stealing

    According to media reports, OpenAI has recently upgraded its security system to deal with the risks of corporate espionage. One of the core measures of this upgrade is to launch the policy of "information isolation", which aims to strictly restrict employees' access to sensitive algorithms and new products. For example, during the development of the o1 model, OpenAI stipulates that only team members who have been strictly reviewed and have obtained the appropriate permissions can discuss relevant details in the public office area. At the same time, the company has also taken a number of physical and network security enhancement measures: store core proprietary technology in isolation

  • What are the major ai models in china? One article to understand the panorama of China's core AI big model

    In the past two years, China's AI big model has developed rapidly, forming an ecological pattern of "general+industry". General big models such as Baidu Wenxin Yiyan and Ali Tongyi Qianwen are outstanding in Chinese understanding and multimodal generation; DeepSeek achieves GPT-4 reasoning ability with 1/10 parameters; Kimi, the dark side of the moon, leads the long text processing with a 200000 Chinese character context window; The GLM-4 of Zhipu AI became the first domestic 100 billion model supporting video dialogue. Professional models such as Baichuan Intelligence have emerged in vertical fields such as medicine and finance, covering more than 1000 hospitals. Open source ecology accelerated the adaptation of domestic chips, and 15 manufacturers completed in-depth optimization. In terms of industrial applications, the efficiency of engineering documents has been improved by 60%, and cases such as 24-hour personified customer service highlight their value. The future trend shows three directions: model distillation, end side deployment and open source ecology. China AI is moving from laboratory to the forefront of industrial reform.

    AI large model AI development in China A word of Wenxin

Hot text

  • 3 days
  • 7 days

Business opportunity of webmaster

advertisement

Business cooperation Infringement complaint Advertising services Copyright Notice recruit

© CopyRight 2002- two thousand and twenty CHINAZ.COM