model

Doubao Work announced the opening of third-party large model services to enterprise clients, with the first batch including models such as Kimi K3.1 Pro from Moon's Dark Side, Zhipu GLM-5.3, and DeepSeek V4.1 Flash. Enterprise users can directly switch models in the Doubao Work dialogue interface, no longer limited to Doubao's own models.Previously, Doubao Work supported users to access third-party models independently, and custom access required users to settle fees separately with the model vendors. The tool invocation and product costs of Doubao Work are still consumed separately. This time, third-party models are directly included in the product's model selection menu, lowering the switching and usage threshold.

According to Cointelegraph, Stellar has launched Project Thruput, a new execution model that has gone live on FutureNet. Early benchmarks show that the model achieves 7.02 million payments per second by scaling across multiple machines.

According to a report by Blue Whale News, at the Tencent Global Digital Ecosystem Conference, Tencent Group Senior Executive Vice President and CEO of the Cloud and Smart Industry Group, Tang Daosheng, mentioned in his speech that they will release the new model Hy4 and Hy4 flash.Compared to the released Hy4 preview, Hy4 enhances the capabilities for various long-range and real productivity tasks, improves the efficiency of the thinking chain, and introduces visual understanding for the first time. Hy4 flash is positioned as a practical model that offers cost-effectiveness.

Alphabet CEO Sundar Pichai stated that the new AI model Carbon is being launched first to cybersecurity partners.

Alphabet CEO Sundar Pichai stated that the new AI model Carbon is being tested with the U.S. government.

Alphabet (GOOG.O) CEO Sundar Pichai stated that the new AI model Carbon is Google's strongest model.

Google (GOOG.O) launched the Gemini 4 Carbon cutting-edge model.

NVIDIA (NVDA.O) launched the Nemotron 3.5 speech transcription model.

B.AI announced a significant upgrade to the model ranking list, adding four major evaluation dimensions beyond real usage popularity: the popular model tracking weekly token usage ratio, reflecting community selection trends; the call ranking supports real-time statistics of token and API call popularity over multiple time periods; the price ranking provides comparisons of input, output, and weighted prices, facilitating precise cost control; the intelligent ranking combines the Artificial Analysis intelligent index with a cutting-edge scatter plot of cost-effectiveness, helping to quickly identify high ROI models.The new ranking list can meet the needs for exploring cutting-edge capabilities, controlling call costs, and optimizing engineering selection.

Bloomberg Intelligence released a report on October 28, stating that by 2030, China will be globally leading in 8 out of 17 key industries, including large language models (LLM), which are core to artificial intelligence. The report, led by Francis Chan, assesses that the U.S. and China are moving towards "AI joint leadership," with neither side likely to achieve a comprehensive advantage.The report lists the 8 areas where China is expected to lead: LLM, optical communication transceiver modules, humanoid robots, general commercial vessels, autonomous and connected vehicles, energy storage, green hydrogen, and biopharmaceutical CDMO. On the hardware side, China currently lags in chip design, foundry, memory, and device materials, but the report predicts that by 2030, it will be globally competitive, although complete self-sufficiency remains challenging. The report also notes that the U.S. may maintain a slight edge in model development, while China holds a favorable position in software optimization and low-cost AI implementation.The report also highlights risks: Chinese AI companies face intense price competition, and profitability will need to wait until pricing strategies become more rational. Additionally, trade barriers and export restrictions are more likely to alter the geographical landscape of China's industrial expansion rather than halt it.

On October 28, B.AI announced that ByteDance's Seedance series video generation models have been integrated into the API, with official support for full access. A total of four models have been launched, covering different tiers from flagship to lightweight.Seedance 2.5 focuses on multimodal audio-visual joint generation, with a maximum single duration of 30 seconds, up to 1080p 10-bit, supporting up to 50 reference materials and targeted timestamp editing; Seedance 2.0 emphasizes high-quality output, supporting up to 4K 10-bit with audio-visual synchronization, and allows control of the first and last frames as well as extensions before and after; Seedance 2.0 Fast balances speed and cost, providing quick generation at 480p/720p, suitable for advertising variants and storyboard iterations; Seedance 2.0 Mini meets budget-sensitive creative and prompt testing needs at a lower unit price. Developers can now call the above models through the B.AI API, and are welcome to log in to chat.b.ai/chat to experience it.

The Science and Technology Innovation Board Daily reporter Huang Xinyi exclusively learned that Baidu plans to disclose the research and development progress and results of its world model in the coming days. The world model is one of the popular directions in the current AI field, focusing on modeling and predicting physical environments. It is reported that if this model is used for autonomous driving, it could enhance prediction and decision-making capabilities in complex scenarios, providing support for the large-scale implementation of driverless cars.In the past year or two, Baidu's autonomous driving service, Luobo Kuaipao, has covered 28 cities worldwide and has accumulated a certain scale of real road data. In recent years, Baidu has also been continuously making strategic layouts in the field of embodied intelligence, serving more than 50 key enterprises in the industrial chain, including the Beijing Humanoid Robot Innovation Center, Zhiyuan, Xinghaitu, and Yushu. As for the specific applications of this model and its relationship with Luobo Kuaipao, it is currently unclear, and Baidu has not commented as of the time of publication.

Market news indicates that Nvidia and Meta will expand their collaboration on MUSE and the MUSE series models against a backdrop of steady growth.

OpenAI recently announced that its AI model has cracked the Exponential Time Hypothesis, abbreviated as ETH, which has been a longstanding conjecture in computer science for over twenty years, and is not related to Ethereum. This hypothesis focuses on how difficult particularly hard problems are, using the Boolean Satisfiability Problem (SAT) as an example: given a set of logical conditions, is there an assignment that makes all conditions true simultaneously? As the problem size increases, the best-known methods essentially can only try possible answers one by one, with the time cost growing exponentially with the size. The Exponential Time Hypothesis further asserts that such problems cannot be significantly accelerated, and there are no truly sub-exponential time efficient algorithms.Since its proposal in 1999, this hypothesis has been a crucial cornerstone of theoretical computer science, with numerous papers proving that certain problems are hard and that some algorithms are close to optimal based on it. If the Exponential Time Hypothesis were to be overturned, it would imply that many problems considered nearly impossible to solve faster might theoretically have better solutions. OpenAI claims that its model has found a deterministic algorithm that can accurately count the number of solutions to the k-SAT problem in sub-exponential time, and it has been formally verified using Lean. If this result withstands expert scrutiny, it will directly challenge the Exponential Time Hypothesis.

According to the report by the Science and Technology Innovation Board Daily, citing the Daily Economic News, the latest data from OpenRouter shows that last week (from October 19 to 25), the total call volume of global AI large models was 188 trillion tokens, a month-on-month increase of 14.63%. The top five global call volumes last week were all Chinese AI large models, with DeepSeek V4.1-Flash-1017 ranking first, having a weekly call volume of 33.6 trillion tokens, a month-on-month increase of 801%.Among the models listed during the same period, the weekly call volume of Chinese AI large models was 93.54 trillion tokens, a month-on-month increase of 18.14%; the weekly call volume of American AI large models was 22.47 trillion tokens, a month-on-month increase of 29.44%. The weekly call volume of Chinese large models has exceeded that of the United States for twenty-six consecutive weeks, ranking first in the world.

OpenAI announced that to prevent AI models from breaching security limits, it has required the most advanced models to be trained in environments where direct access to the live internet is not possible. Previously, an internal research model had gained limited internet access through a network configuration vulnerability, prompting the company to suspend training.After security rectifications, OpenAI resumed training on October 1, but still restricts internet access.

On October 23, B.AI announced the domestic model brand Token weekly ranking from October 16 to 22: DeepSeek continues to lead, with Xiaomi MiMo closely following, with only a difference of 16.54 B Tokens. Xiaomi has seen growth for 4 consecutive weeks, and the platform Token usage ratio has increased by 8.13 percentage points.The competition in the mid-tier is heating up, with MiniMax experiencing a month-on-month increase of over 5,200%, becoming this week's growth dark horse. Qwen saw a month-on-month growth of 80.86%, narrowing the gap with the third-place Zai to 0.18%. Popular model discounts are ongoing: DeepSeek-V4.1-Flash-1017 enjoys a 40% discount, Qwen3.8-Flash enjoys a 20% discount, and GLM-5.3-Flash enjoys a 30% discount. B.AI continues to gather high-quality models, helping developers unlock diverse AI capabilities at low costs. Welcome to log in to b.ai to explore.