Baidu

On 4 September, according to information from sources quoted by AI Front, the former DeepSeek researcher, Wei, recently led a team to the Department for the Development of 100-degree Basic Models, where he was the head of the MMA. He joined 100 degrees in the first half of this year, before he was primarily responsible for research and development in multi-modular directions such as OCR. Wei is the first author of DeepSeek-OCR: Contexts Optifical Corporation and DeepSeek-OCR 2: Vision Capital Flow. Two..

On 2 July, according to LateLatePost, Sun Tianxiang, a technical expert in the field of large models, has recently officially joined 100 degrees as head of the BMU. At the same time, Sun Tianxiang entered the 100-degree Model Committee (BMC). Sun Tianxiang, a post-95 model expert, has accumulated a great deal in the academic and industrial fields. During the doctorate, he was the leader of the ChetGPT open source model MOSS, who had published more than 40 papers at top-level meetings in ICML, ICLR, NeuurIPS and so on..

On May 9th, 100% of the news released a new generation of large-scale basic models - the Mandarin model 5.1. At present, Mansion 5.1 is on line in the 100-degree thousand-show model square and the Mansion Web, open to business users and developers. According to the 100-degree official presentation, the model uses “multi-dimensional elastic pre-training” technology, with only the pre-training costs of the industry-size model of approximately 6%, leading to basic results and making it to LMArena's first and fourth in the world. 5.1 The significant increase in integrated capabilities benefits from key technologies such as “multi-dimensional elastic pre-training”. The..

According to March 13, yesterday, OpenClaw's father, Peter Steinberger, accused X of accusing Qing of extracting skill data from ClawHub and importing his home, SkillHub, triggering a continuing debate about the "open spirit" and the "boundary of big factory behavior". At first, @Steinberger, an Internet friend, said that the Quest in his home, SkillHub, "takes all the skills of ClawHub." Steinberger then replied that he had received mail complaining about..

ON NOVEMBER 14, A KEYNOTE SPEECH ENTITLED "EMERGING EFFECTS" WAS DELIVERED BY THE FOUNDER AND CEO OF THE 100-DEGREE WORLD CONGRESS IN BEIJING YESTERDAY. HE STRESSED THAT WHEN AI CAPABILITIES ARE INTERNALIZED INTO THE ORIGINAL CAPACITY OF ENTERPRISES AND INDIVIDUALS, INTELLIGENCE IS NO LONGER A COST BUT A PRODUCTIVITY. HE POINTED OUT THAT AI IS UNDERGOING AN IMPORTANT TRANSFORMATION IN ITS INDUSTRIAL STRUCTURE, MOVING FROM A PREVIOUSLY UNHEALTHY “PROPER PYRAMID” TO A MORE SUSTAINABLE “VERTED PYRAMID” MODEL. LEE YON-HON EXPLAINS THAT, IN THE TRADITIONAL STRUCTURE OF THE PYRAMID, THE CHIP MANUFACTURER OCCUPIES MOST OF THE VALUE AND IS LOCATED ON IT..

ON NOVEMBER 13TH, AT TODAY'S 2025-DEGREE WORLD CONGRESS, THE 100-DEGREE FOUNDER, LI JIANG HONG, STATED THAT THE VAST MAJORITY OF THE SEARCH RESULTS WERE GENERATED BY AI, WHILE THE FIRST RESULT HAD A MEDIA COVERAGE OF 701 TP3T. "YOU NOW SEARCH FOR 10 QUESTIONS, 7 ANSWERS ARE RICH MEDIA, NOT A TEXT LINK, BUT A PICTURE, A VIDEO, A LIVE BROADCAST, EVEN A DIGITAL PERSON." ACCORDING TO LI, 100 DEGREES IS THE MOST RADICAL OF ALL SEARCH ENGINES AROUND THE WORLD. RECONSTRUCT THE SEARCH RESULTS PAGE WITH 100 DEGREES OF AI, NOT SIMPLY SEARCHING FOR KNOTS..

ON NOVEMBER 13TH, DURING TODAY'S 2025-DEGREE WORLD CONGRESS, THE 100-DEGREE FOUNDER, LI JIANGHONG, OFFICIALLY LAUNCHED THE MAGNA CARTA MODEL 5.0. ACCORDING TO LI, THE MAGNA CARTA MODEL 5.0 HAS BEEN UPGRADED IN TERMS OF TOTAL MODULAR UNDERSTANDING, CREATIVE WRITING, INTELLIGENT BODY PLANNING, AND COMMAND COMPLIANCE. SPECIFICALLY, THE MACRO MANSION MODEL 5.0 IS A NEW GENERATION OF WHOLE-STATE MODELS THAT SUPPORT THE INTEGRATION OF UNDERSTANDING AND GENERATION, USING A UNIFIED SELF-RETURN STRUCTURE. 1AI WAS INFORMED BY THE GENERAL ASSEMBLY THAT THE MANDARIN LARGE MODEL 5.0 USES A SUPER-SORTED HYBRID EXPERT STRUCTURE WITH PARAMETERS OF OVER 2.4 TRILLION. MAN-SAMA..

On November 9th, 100-degree government announced that its new generation of mega-models, ERNIE-5.0-Preview-1022, outnumbered GPT-5-High, Qwen3-Max, DeepSeek-R1-0528 on the LMArena ranking list, ranked second and highest in the world. However, hundreds of times did not disclose more information about ERNIE-5.0-Preview-1022, but indicated that it was planned to be officially published in the near future. 1AI follow-up will remain focused. It means the model is in creative writing,..

SEPTEMBER 9, NEWS. TODAY, THE WAVE SUMMIT IN-DEPTH LEARNING DEVELOPER CONFERENCE 2025, BEIJING. THE 100-DEGREE CHIEF TECHNICAL OFFICER AND DIRECTOR OF THE CENTRE FOR ADVANCED LEARNING TECHNOLOGY AND APPLIED NATIONAL ENGINEERING RESEARCH OFFICIALLY RELEASED THE MAGNA CARTA MODEL X1.1 DEEP THOUGHT MODEL. IT WAS DESCRIBED THAT THE X1-DEEP THINKING MODEL WAS BASED ON THE IN-DEEP THINKING MODEL BASED ON G4.5 TRAINING AND X1.1 WAS UPGRADED AGAIN. THE MODEL IS SIGNIFICANTLY ENHANCED IN TERMS OF DE FACTO, COMMAND COMPLIANCE, INTELLIGENCE, ETC. 1AI NOTES THAT USERS ARE ALREADY ABLE TO..

August 19 news, in yesterday's Baidu AI Day open day, Baidu library joint Baidu net disk released the world's first full-end universal intelligent body "GenFlow2.0". The product supports more than 100 expert intelligences to work at the same time, claiming that 3 minutes to complete more than 5 complex tasks in parallel, generating speed 10 times faster than the mainstream products of the same type, to achieve "minute-level delivery, the process can be intervened in, the memory can be traced". According to the introduction, GenFlow2.0 has been formally launched on Baidu Wikipedia Web, App, without queuing, invitation code, ...

August 12, Baidu announced yesterday that Baidu Search PC home page is fully online series of AI functions, upgrading the traditional information portal into a task hub. The upgraded home page has a new "super intelligent two-line box", which supports one-click in-depth search and can answer complex questions; it also introduces a "workbench" module, which integrates three major tools, namely AI Reading, AI Writing and AI PPT. AI Reading can summarize links, pictures and oversized documents with one click; AI Writing provides more than 20 styles of writing and supports underlining modification; AI PPT can automatically generate presentations based on the whole network....

Aug. 7 (Bloomberg) -- Baidu plans to launch a new inference model capable of handling more complex tasks by the end of this month (the end of August 2025) in order to gain a foothold against the likes of DeepSeek and OpenAI, according to a report in today's Wall Street Journal by people with knowledge of the situation. Baidu also plans to release an updated version of its core foundation model, called Ernie 5.0 (note: Ernie is Wenxin), in the coming months, the report added. Baidu released the "strongest" Wenxin model in March ...

July 2 news, Baidu commercial R & D team released its own video generation model "MuseSteamer" and creation platform "Painted Thinking", MuseSteamer is the world's first to realize the integration of Chinese audio and video video generation video model. MuseSteamer is the world's first video model that realizes the integrated generation of Chinese audio and video. The technology can realize the synergistic creation of picture, sound effects and human voice lines, breaking the fragmented process of traditional AIGC video, which is "picture first and then dubbing". It is reported that MuseSteamer tops the world's first in the authoritative list VBench I2V with a total score of 89.38%, and supports the generation of a 10-second 1...

June 30, Baidu today officially open source Wenxin Big Model 4.5 series of models, covering 47B, 3B activation parameters of the Mixed Expert (MoE) model, and 0.3B parameters of the dense model and other 10 models, and realize the pre-training weights and reasoning code completely open source. At present, Wuxin Big Model 4.5 open source series can be downloaded and deployed on platforms such as Flying Paddle Star River Community and HuggingFace, and the open source model API service can also be used on Baidu Intelligent Cloud Qianfan Big Model Platform. Wuxin Big Model 4.5 was released in ...

May 22 news, according to surging news reports, in the first quarter earnings call, Baidu founder Robin Li said, Baidu from a strategic point of view, the model research and development focus on areas with practical application value. Robin Li said, "In the past few months, the speed of iteration of the basic model appeared to accelerate the trend, one thing is always clear, the real value of the basic model is the application built on the model, the basic model field is very broad, so we do not have to take the lead in every direction." Regarding the development of driverless cars, Robin Li said, "We are open to cooperation, and are actively working with online car rental platforms, car...

Heartbeat is a new Baidu super intelligent product for cell phones, with "AI Task Completion Engine" as the core, Heartbeat can help users realize complex task disassembly, dynamic execution and visualization of results delivery through natural language interaction. Relying on the large model and multi-intelligence body synergy, Heartbeat is committed to becoming the user's "super brain" + "strongest assistant", so that the user is liberated from the cumbersome process, one-stop intelligent decision-making for complex problems, efficiency scheduling, and realize the whole process of hosting. Heartbeat can be applied to in-depth knowledge analysis, travel planning, learning office and other core life scenarios. Heartbeat Features Mastermind scheduling system (task full process...

April 25 news, today Create2025 Baidu AI developer conference was held in Wuhan, at which Baidu officially launched a general multi-intelligence body collaboration App heart sound. At present, the product has been launched on the Android version, iOS is on the shelf. At present, the heart ring App has been online, including routine tasks, city tourism, AI dating, in-depth research, legal counseling, health advice, wisdom charts, test question explanation, AI picture books, game development, including the top ten scenarios of more than 200 types of tasks, the future plans to expand the types of tasks to more than 100,000. According to 1A...

April 25 news, in today's Baidu Create developer conference, Baidu founder Robin Li released the Wenxin big model 4.5 Turbo, which compared to the Wenxin 4.5 faster, the price dropped by 80%, the input price per million tokens is only 0.8 yuan, the output price is 3.2 yuan, only 40% for DeepSeek-V3. 1AI learned from the official introduction that the two new models have three major features: multimodal, strong inference, and low cost. Wenxin 4.5 Turbo in multimodal capabilities, de-illusionization, logic...

April 10 news, 1AI from Baidu official learned, Wenxin big model 4.5 Turbo will be officially unveiled on April 25 Create conference, but the official did not describe in detail this new model will bring upgrades. Create 2025 Baidu AI developer conference will be held on April 25 at the Wuhan Sports Center, the conference will be around the MCP, DeepSeek, Agent, AI programming and other hot topics set up six sessions. In addition, Baidu also said it will bring the main...

March 31 news, in today's Baidu AI DAY, Baidu released the first end-to-end speech language big model based on the new mutual attention (Cross-Attention), announced the realization of ultra-low latency and ultra-low cost, in the voice Q&A scenario of the telephone voice channel, the cost of invocation decreased by about 50%-90% compared with the average value of the industry. On the same day, the Wen XiaoYin On the same day, WenxiaoYin announced a brand refresh and took the lead in accessing the model, as well as bringing upgrades to functions such as multi-model fusion scheduling and picture Q&A. After accessing the model, Wen Xiaoyan can not only support a more realistic language chat effect, but also support Chongqing, Guangxi,...

March 16th news, Baidu today officially released Wenxin Big Model 4.5 and Wenxin Big Model X1, two models have been online in Wenxin Yiyin official website, free of charge to users. Wenxin Big Model 4.5 has been launched on the Baidu Intelligent Cloud Thousand Sails Big Model platform, enterprise users and developers can log in to call the API; Wenxin Big Model X1 will soon be launched on Thousand Sails, Baidu Search, Wen Xiaoyin App and other products will be accessed Wenxin Big Model 4.5 and Wenxin Big Model X1. According to the 1AI understanding, the Wenxin Big Model 4.5 is the first Baidu native multi-modal Big Model.... ...

February 28 news, Baidu today announced that Wenxin big model 4.5 will be officially released on March 16th. Baidu officials said, Wenxin big model 4.5 not only in the basic model capability has been greatly improved, and has the ability to native multimodal, deep thinking and other capabilities. According to a summary of 1AI's previous report, Baidu said earlier this month that the Wenxin 4.5 series will be gradually launched in the next few months, and plans to officially open source from June 30th. Baidu also previously revealed that with the iterative upgrading of Wenxin's large model and the decreasing cost, Wenxin Yiyin will be released on April ...

February 27, Baidu plans to release its next-generation artificial intelligence model, Wenxin 4.5, in mid-March, according to Reuters, citing sources familiar with the matter, and the new version will achieve improved capabilities in a number of areas, including reasoning. The upgraded Wenxin 4.5 will have stronger multimodal capabilities, the people familiar with the matter said. The multimodal AI system is capable of processing and integrating different types of data, such as text, video, images and audio, and converting between different formats. Baidu said earlier this month that the Wenshin 4.5 series will be rolled out gradually over the next few months and is scheduled for June 30...

February 21, 2011 - Baidu Search PC has now launched DeepSeek full-blooded version and provides networking services, which can be experienced by clicking on "AI" at the bottom of the homepage search box. 1AI noted that Baidu Search has accessed the full-blooded version of DeepSeek-R1, which integrates Baidu's networked search function and has technical capabilities such as retrieval-enhanced RAG. Baidu Search and Wenshin Intelligent Body Platform announced on February 16th that they have fully access to the latest deep search functions of DeepSeek and Wenshin Big Model.