ByteDance

On 4 December, the byte-jumping bean bag team, in collaboration with China, released a technological preview of the soybean cell phone assistant, which resulted in feedback from users on a pilot project for Nubia M153, which resulted in irregular access to Twitter. Recently, the byte-jumping bean bag team, in collaboration with China, released a technological preview of the bean bag phone assistant, which resulted in feedback from users on a pilot project for Nubia M153, which resulted in irregular access to Twitter. In this case, the soybean team sent a letter and responded last night. The bean bag said that its follow-up was offline with the cell phone assistant's ability to operate the Twitter..

On November 15th, the news that recently, byte beats the launch of Universal AI Smart "Lumine" shows a high degree of autonomy and cross-scenes in the Open World Game "The Gods." The system is known to be able to perform several hours-long lead line tasks without manual intervention and to achieve autonomous navigation in unfamiliar scenes. Based on the Qwen2-VL-7B-Base model, the research team integrated senses, reasoning and operations using a human interactive paradigm, creating the "Base" "Instract" "Thinking" level three..

In the news of September 22, the byte-jumping volcano engine recently released its brand-new generic translation large model, Doubao-Seed-Translation, which supports 28 languages of translation into one another, covering the main languages of China, E, Japan, Korea, Germany, France, S, Russian and other languages. Officially, the bean-bag translation model supports the translation of dozens of languages into one another, providing faithful, tunneled and fluent translations, with Chinese and English translations approaching Deepseek-R1, and national multilingual translations exceeding or evening GPT-4o / Gemini-2...

September 8 news, byte jump that dream AI announced on September 5, that dream picture 4.0 officially released. Dream Picture 4.0 supports multimodal picture generation for the first time, and the same model realizes text-generated pictures, image editing, and group picture generation. The same model can realize text-to-picture, image editing, and group picture generation. It can flexibly control the details of the picture through natural language. Text Generation: Stronger command compliance, higher resolution support, faster generation speed Image Editing: Input single/multiple images + natural language commands, you can make any kind of editing modifications Group Image Generation: Generate multiple images with correlation at one time, a good helper for your inspirational brainstorm...

On July 22, the Seed team of ByteDance announced its latest achievement: a generalizable large-scale model of robot manipulation "Seed GR-3" that supports long sequences of complex manipulation tasks. According to the introduction, GR-3 is a large-scale vision-language-action (VLA) model. It shows good generalization ability to new objects, new environments, and new commands with abstract concepts. In addition, GR-3 supports efficient fine-tuning of a small amount of human trajectory data for fast and cost-effective adaptation to new tasks.GR-3 is useful in dealing with long-period and dexterity tasks (including tasks that require two-handed manipulation and ...

July 7 news, byte jump's AI native integrated development environment (IDE) Trae announced on July 4 that it officially open-sourced its core component, Trae-Agent. Trae officials also said that they are looking for active users and intelligent body developers who are willing to contribute to the construction of an open ecosystem of intelligent bodies. According to the GitHub page, Trae Agent is an intelligent body based on the LLM generic software engineering task. It provides a CLI interface that understands natural language commands and uses various...

June 27 news, according to the "new wisdom core" today broke the news, byte's Seed team will recruit a number of robotics-related business No. 1 position, including robotics product leader, robotics engineering and technology leader, as well as the person in charge of the big model of embodied intelligence. This series of actions shows that Seed's determination to accelerate the layout in the field of robotics and embodied intelligence. Breaking news mentioned, byte jumping robotics team leader Kong Tao has left in early June this year, the Seed team's recruitment of a number of key positions, or with the team structure adjustment and business strategy shift. According to surging news reports, byte...

June 6 news, byte jumping Seed team today announced the release of image editing model SeedEdit 3.0, has been opened in the namely dream web end of the test, beanbag App is also about to go online. The need to rely on AI to complete command image editing is widely present in visual content creative work. However, previously, the image editing model was relatively limited in its ability to maintain the subject & background and follow instructions, resulting in a low usability rate for editing images. According to the official introduction of ByteDance, SeedEdit 3.0 is based on the literate image model Seedre...

May 28 news, according to the first financial report today, byte jumping security and risk control department issued an email saying that, for the sake of preventing the risk of data leakage, since June 30, will be within the part of the batch to disable third-party AI development software, including the AI programming tools Cursor, Windsurf, etc., and will be byte of programming assistant Trae as an alternative program. According to 1AI's report in March this year, Byte launched the domestic version of Trae, saying it is "the first domestic AI native integrated development environment tool (AI IDE)", configuring D...

On May 11, ByteDance's technical team announced that DeerFlow, a new Deep Research project based on LangStack, was officially open-sourced. According to the official introduction, DeerFlow supports functions including deep research, MCP integration, report AI-enhanced editing, and podcast generation. It is worth mentioning that DeerFlow supports the popular Replay mode (i.e., the process of multiple rounds of interaction with a large model is restored in the form of fast playback). DeerFlow utilizes a new ...

April 29, 2011 - Doubao's video generation function has recently been upgraded, with significant improvements in semantic comprehension, shadow quality, style type, and motion consistency compared to the previous version, according to officials. The upgrade is supported by Doubao's video generation model - Seedance, which can be experienced by opening the Doubao App dialog box and selecting "Move Photos", or selecting "Video Generation" in Doubao's computerized version and doubao.com webpage, and typing in the text or uploading a reference image. or upload a reference image to experience it. 1AI notes that earlier this month, ByteDance released Doubao 1.5 Deep Thinking Model, which...

On April 27th, ByteDance Seed officially launched the 2026 Top Seed School Recruitment Program for Top Model Talents. Previously, Top Seed opened the recruitment of research interns in March, and today it officially launches the recruitment program for the 2026 class, with a plan to recruit about 30 top PhD candidates. The research topics of this year's Top Seed include large language models, machine learning algorithms and systems, multimodal generation, multimodal understanding, speech, etc., which basically cover all areas of large model research. It is worth mentioning that this year's Top Seed ...

April 19, 2011 - Seedream 3.0, a new text-generated image model from ByteDance, has excelled in both internal and external reviews, surpassing its predecessor, Seedream 2.0, and competing with mainstream systems such as GPT-4o, Midjourney v6.1, and Imagen 3, among others. The model doubles the amount of training data, adds pre-processed masked defective images, and employs new techniques such as resolution-adaptive sampling and mixed-resolution training to ensure high-fidelity output for images of different sizes. Seedr...

April 17 news, in today's volcano engine AI innovation tour exhibition Hangzhou station site, byte jumping's volcano engine president Tan to be released the latest beanbag 1.5・deep thinking model. According to reports, the model has outstanding performance in professional fields such as math, programming, scientific reasoning, and general tasks such as creative writing. Mathematical reasoning AIME 2024 test scores tie OpenAI o3-mini-high, programming competition and scientific reasoning test scores are also close to o1. The model also demonstrates excellent generalization ability in general-purpose tasks such as creative writing and humanities quiz...

March 28 news, today, byte jumping's bean bag announced that a new version of the depth thinking function opened for testing. 1AI learned from the official introduction that, unlike the previous deep thinking, the new version of the function combines the chain of thought of the reasoning process with the depth of the search, and supports searching while thinking. During the thinking process, the beanbag can call tools and search information multiple times based on reasoning to provide more comprehensive and accurate results. Developing Programs and Planning Beanbag can reason about missing information based on the thinking results and continue to directional search to make the final results more complete. Taking Qingming travel as an example, after 3 rounds of search, Beanbag synthesizes all kinds of situations and writes a complete enough...

March 6 news, last Tuesday a netizen found that byte jumping's AI assistant beanbag is testing deep thinking model in a small scale, but the access is not DeepSeek model. On March 5, 2025, Doubao announced the launch of the "Deep Thinking" reasoning mode, which comprehensively covers Q&A, search, writing and reading scenarios. To access the DeepSeek reasoning mode, users only need to turn on the "DeepSeek Beta" switch in the input box. During last week's test, the person in charge of Doubao responded that the current test is a different experimental version of its own Deep Thinking model. Also...

March 5 news, 1AI learned from byte jumping, that dream AI today on-line "action imitation" function, the user from the "digital person" entrance to enter, just upload a picture of a character and a reference video, you can generate a dynamic video, so that the picture of the character simulation reference video character action can also realize the emotion of a one-to-one reproduction. Users can upload a picture of a character and a reference video to generate a dynamic video, allowing the character in the picture to simulate the movements of the character in the reference video, and also realizing one-to-one restoration of emotions. The feature supports different frame sizes, including portrait, half-body and full-body, and is powered by ByteDance's Intelligent Creation Digital People team. The team uses a mix of explicit and implicit feature-driven approaches to synchronize body movements and facial expressions across all frames...

March 3 news, byte jumping today announced that "China's first AI native integrated development environment (AI IDE)" Trae domestic version is officially online, configured with Doubao-1.5-pro, and support for switching full-blooded version of DeepSeek R1, V3 models. Trae was previously released overseas, with built-in GPT-4o and Claude-3.5-Sonnet models. The domestic version carries different models, but the functions are basically the same. Trae is positioned as an "intelligent collaborative AI IDE", supporting code completion...

February 26th, according to "LatePost" reported yesterday, after the flying book, byte jumping AI video generation product "namely dream" is considering to use DeepSeek. The explosion of DeepSeek also brings the demand overflow and linkage play. and linkage play. Reportedly, for example, first use DeepSeek to generate more detailed video scripts, and then generate videos in that dream. The number of users of DeepSeek grew rapidly around the Spring Festival. According to QuestMobile, a third-party monitoring platform, DeepSeek's weekly activity at the end of December last year was about 760,000, and as of 2...

February 17 news, byte jump in January this year released a new AI programming product for developers Trae, which is an AI Chinese IDE tool, but at that time only Mac version. According to the feedback of 1AI partners, the tool is now online Windows version, supporting Windows 10 and Windows 11 systems. Trae is powered by SPRING PTE, a Singaporean company under ByteDance, and supports AI Q&A, code auto-completion, Agent-based ...

February 12, 2011 - The Beanbag Big Model team announced today that the Byte Jump Beanbag Big Model team has proposed a new sparse model architecture, UltraMem, which effectively solves the problem of high access memory in MoE inference, improves inference speed by 2-6 times compared to the MoE architecture, and reduces the cost of inference by up to 83%. The study also reveals the Scaling Law of the new architecture, proving that it not only has excellent Scaling characteristics, but also outperforms MoE in terms of performance. The study also reveals the Scaling Law of the new architecture, demonstrating that it not only has excellent scaling characteristics, but also outperforms MoE. Experimental results show that the training scale of 20 million value...

January 22 news, byte jump's Doubao big model 1.5 Pro officially released today, the new Doubao-1.5-pro model enhanced comprehensive ability, in the knowledge, code, reasoning, Chinese and other benchmarks, the comprehensive score is better than the GPT-4o, Claude 3.5 Sonnet and other models. At present, Doubao-1.5-pro has been grayed out on line in Doubao App, while developers can also directly call the API in the volcano engine. According to the official introduction of the byte jumping, Doubao big model 1.5 P...

ByteDance recently launched Trae, an AI integrated development environment (IDE) for Chinese developers, which aims to compete with internationally recognized tools such as Cursor and Windsurf, especially in terms of the experience of Chinese developers, and hopes to fill the shortcomings of existing tools. In the daily programming, many developers often need to switch between English and Chinese, and although this mixed working mode is commonplace, it is inconvenient for many Chinese developers. Existing IDEs such as Cursor, Windsurf and ...

January 17 news, byte jumping beanbag January 16 announced that beanbag computer version and web version of the new online AI programming function. This function supports uploading multiple local code files with one click, introducing GitHub open source repository in real time, and quickly obtaining the complete context of the project without copying the code segment by segment. 1AI learned from the official introduction that Doubao's new AI programming function is equipped with a brand new code editor that supports immersive reading; code snippets that need to be interpreted or adjusted can be accurately circled; and the directory can be easily switched when viewing the code repository. In addition, users can also by beanbag to...