NAll news

All news

Wednesday626 items

  1. QbitAIModel releases
    AI score56

    Ant Ling Open-Sources Ling-3.1-flash-1019, Intelligence Index Rises to 44

    Ling, the large model under Ant Group, has open-sourced Ling-3.1-flash-1019, which scored 44 on the Artificial Analysis Intelligence Index v4.3.2, a 3-point improvement over the previous version. This version was further trained and optimized on the basis of Ling-3.1-flash, and the base model Ling-3.1-flash-base was also open-sourced, laying the foundation for the knowledge, reasoning, and computational efficiency required for long-horizon tasks.

    Same story · 3 reports

    #Model releases#Open-source ecosystem#Agents

  2. Sarvam AI BlogFirst-party
    AI score50

    Sarvam Signs AI Cooperation Memorandum with India's Headquarters Integrated Defence Staff (HQ IDS)

    Sarvam has signed a memorandum of understanding with the Headquarters Integrated Defence Staff (HQ IDS) under India's Ministry of Defence, establishing a unified AI program framework for the Indian Army, Navy, Air Force, and joint tri-service institutions. The three services will share the same set of foundation models, agent infrastructure, and geospatial and perception systems, with Sarvam responsible for designing, developing, and deploying AI applications and providing ongoing technical support. The related work is undertaken by Chanakya, Sarvam's isolated division focused on national security and defense, using an AI technology stack that includes Sarvam 105B, with emphasis on decision support, target acquisition, and data fusion for the sensor-to-shooter chain, while human oversight is retained for operational decisions.

    Sarvam and HQ IDS establish joint framework for AI adoption across the Army, Navy and Air Force

    Same story · 3 reports

    #Industry news#English text is being prepared.#Agents

  3. cnBeta
    AI score40

    Bloomberg Intelligence: China's AI large models will lead the world by 2030

    Bloomberg Intelligence predicts that by the end of 2030, China will lead in 8 of 17 key industries, including large language models. The report says U.S. models may maintain a slight lead at the frontier, while China's expertise in software optimization puts it in an advantageous position for low-cost AI applications. China currently lags in most AI hardware categories, from chips to optical network components, but will be competitive by 2030.

    Same story · 2 reports

    #Industry news#English text is being prepared.

  4. cnBeta
    AI score35

    DeepSeek V5 series appears in official code repository, placeholders already in place before release

    Placeholder code such as DeepSeek-V5, V5-open, and V5-next has appeared in the official DeepSeek Harness code repository, meaning the release mechanism for the V5 series is ready; in the future, the service side can go live by simply changing /models, with no need for a client release. Among these, V5-open follows a path with no version threshold, while V5-next requires a client upgrade, covering both distribution paths. The code does not mean V5 will be released soon, but groundwork for the release has begun.

    #Product updates#DeepSeek

  5. cnBeta
    AI score70

    OpenAI plans to procure Google TPUs

    OpenAI plans to procure Google TPUs as a source of compute beyond Nvidia's AI chips. Google began selling TPUs directly to some customers this year, and Oracle also announced on Tuesday that it would for the first time offer Google TPU-based compute to customers on its cloud platform. OpenAI had previously announced plans to deploy in-house chips later this year; its internal chip Jalapeno, developed in partnership with Broadcom, is led by Richard Ho, who previously participated in Google TPU development.

    Same story · 6 reports

    #Industry news#OpenAI#Google

  6. cnBeta
    AI score44

    AI Drives Up Memory Costs, HP and Other Major PC Makers Shift Some Production Back to China

    Amid a severe shortage of memory and storage components caused by AI demand and rising bill-of-materials costs, PC manufacturers such as HP are moving some production back to China and turning to more competitively priced Chinese contract manufacturers. Dell, the world's third-largest PC maker, is also evaluating similar adjustments, while Acer says it will dynamically adjust its supply chain. Chip industry executives say that if the cost environment does not improve, PC makers may further shift production back to China.

    #Industry news#English text is being prepared.

  7. indigo
    AI score31

    Vals AI estimates that Anthropic will achieve full RSI in July-August 2027

    According to the latest estimate from Vals AI, Anthropic will reach full recursive self-improvement (RSI) around July-August 2027. Rayan Krishnan said that the agentic tasks collected by its RSI Index correspond to the work of AI lab researchers on pre-training, post-training, and framework-level engineering, and that the steepest growth curve is Gemini's.

    #English text is being prepared.#Reasoning#Agents

  8. Peter Yang
    AI score17

    Peter Yang lists four hiring practices that should change

    Commenting on the Anthropic team's decision to cancel coding interviews, Peter Yang listed other hiring practices that need to change: screening candidates by years of work experience or company name, emphasizing resumes and cover letters over actual projects, treating headcount managed as an achievement in itself, and replacing short work trials with endless rounds of interviews.

    Quoted postFelix Rieseberg · @felixrieseberg

    The coding interview is dead, at least for my team at Anthropic. I've seen enough to know that it no longer provides signal. I removed it from loops this month. I have a spicy, optimistic take - I think there's an opportunity for junior engineers: https://t.co/PVvpQCbQtK

    #English text is being prepared.

  9. Elon Musk
    AI score21

    Elon Musk reposted the Grok Imagine prompting guide

    The official Grok Imagine account announced that the Prompting Guide is now live, saying the team is publicly producing the Odyssey series with a small group of creators and inviting users to learn their workflow. Elon Musk reposted the message, adding only the words "Grok Imagine."

    Quoted postGrok Imagine · @imagine

    Our Prompting Guide is live. We're building the Odyssey episodes in public with a small crew of creatives. Learn from our workflows! https://t.co/rt1EYu2ylX

    #English text is being prepared.

  11. ModelBest OpenBMB
    AI score37

    ModelBest focuses on the ForgeKernel kernel optimization framework

    Yuxuan Li announced that, based on previous findings from JustGoal, they redesigned their approach to kernel optimization and launched a self-evolving kernel optimization framework called ForgeKernel (part of the Forge series). It ranked first on both NVIDIA's SOL-ExecBench (FlashInfer-Bench) and Huawei's CANN-Bench, and the paper will be released soon. ModelBest OpenBMB reposted it and said they look forward to the upcoming paper and more results from the Forge series.

    Quoted postYuxuan Li · @_LiYuXuan_

    Based on our previous discovery on JustGoal, we redesigned how we do kernel optimisation and launched a self-evolving kernel optimisation harness called ForgeKernel, a part of our Forge series. We tried ForgeKernel on SOL-ExecBench(FlashInfer-Bench) from NVIDIA and CANN-Bench from HUAWEI. It performed well and easily reached #1 in every list we tested. Here are the ranks: https://t.co/IoM3k9UY19 https://t.co/mHD2RF3qkZ We'll release another paper on ForgeKernel soon.

    Same story · 2 reports

    #Industry news#Agents

  12. ModelBest OpenBMBOpinion
    AI score55

    OpenBMB responds to third-party comparative testing of MiniCPM-OCR

    OpenBMB reposted and thanked @wikiwayne for the comparative test of MiniCPM-OCR, stating that the results and caveats were made public. OpenBMB said it looks forward to the performance after the other party completes local testing on the GX10, and that such real feedback will drive the team to keep improving. The quoted tweet introduces MiniCPM-OCR 1.0 as a 1.3-billion-parameter open-source OCR model, with the accompanying MiniCPM-OCR Jev used to evaluate result reliability, and an official benchmark score of 93.68 on Real5-OmniDocBench.

    Quoted postGuillermo Casaus · @_guillecasaus

    Acaban de lanzar un modelo OCR open-source de solo 1.300 millones de parámetros. Se llama MiniCPM-OCR 1.0 y viene acompañado de MiniCPM-OCR Jev, otro modelo que evalúa la fiabilidad de los resultados. Ambos desarrollados por OpenBMB y ModelBest. Entre ambos permiten: → Extraer texto, tablas y fórmulas de documentos → Procesar PDFs y páginas escaneadas o fotografiadas → Trabajar con documentos inclinados o curvados → Incluir una puntuación de confianza en los resultados para facilitar su revisión manual → Ejecutarse con vLLM, SGLang y Ollama MiniCPM-OCR 1.0 ha conseguido 93,68 puntos en Real5-OmniDocBench, según sus benchmarks oficiales. Pero lo más interesante es Jev. En lugar de corregir los errores automáticamente, evalúa la fiabilidad de los resultados extraídos e incluye una puntuación de confianza para facilitar su revisión. Además, MiniCPM-OCR 1.0 incorpora KIE (Key Information Extraction), una funcionalidad para extraer información clave de los documentos. Y sí, los pesos de ambos modelos son abiertos. 🔗 GitHub: https://t.co/38R4WnytR6 🤗 Hugging Face: https://t.co/57lvy4gscl Más recursos y demo en el primer comentario 👇

    #Industry news#Multimodal#Evaluations / benchmarks

  13. Leiphone
    AI score24

    BAAI's SoulAgent Enters RDFZ, Expanding Educational Application Scenarios

    The BAAI's SoulAgent visited the High School Affiliated to Renmin University of China (RDFZ) from October 19 to 24, carrying out application practices around scenarios such as "Help Me Listen to Class," teacher AI avatars, essay grading, and teaching Q&A. On October 21, 27 open classes were connected to "Help Me Listen to Class," using AI quick notes, intelligent summaries, and continuous Q&A to assist teachers in reviewing classes and teaching research debriefs. SoulAgent is now open for registration and can be experienced via the official website soulagent.world or the WeChat mini program.

    #Product updates#Agents#Tutorials / practice

  14. Leiphone
    AI score17

    Narwal at Ten Years: From the J1's Automatic Mop Washing to a Global AI Smart Cleaning Expert

    Ten years after its founding, Narwal has established four product lines—robot vacuums, wet dry vacuums, vacuum cleaners, and mite removers—with cumulative patent applications exceeding 1,500. Its products have entered more than 30 countries and regions worldwide and served over 6 million user households. On Double 11 in 2019, its first product, the J1, launched as the world's first robot vacuum capable of automatically washing its mop. Subsequently, the J2 introduced modular automatic water supply and drainage, the J3 featured the DirtSense™ wastewater recognition system, and the J4 used a cyclonic-guided anti-tangle roller brush to address hair tangling. On the occasion of its tenth anniversary, Narwal has clearly defined its brand positioning as a global AI smart cleaning expert.

    #Industry news

  15. Shao Meng
    AI score49

    Anthropic team drops coding interviews in favor of weekend prototypes

    Felix Rieseberg of the Anthropic team said that this month they have removed the coding interview from the hiring process, because small React/algorithm questions cannot measure real impact, and Claude is stronger at such questions. Instead, he assesses business impact, radical ideas, and runnable weekend prototypes, and believes that the "hunger" of junior engineers can outweigh seniority. Team newcomers Anthony Morris and Alice Zhao built a Claude Code graphical interface prototype over the weekend, and a public version was available about eight days later.

    Quoted postFelix Rieseberg · @felixrieseberg

    The coding interview is dead, at least for my team at Anthropic. I've seen enough to know that it no longer provides signal. I removed it from loops this month. I have a spicy, optimistic take - I think there's an opportunity for junior engineers: https://t.co/PVvpQCbQtK

    Same story · 5 reports

    #Industry news#Anthropic#English text is being prepared.

  17. Peter Yang
    AI score22

    Peter Yang Mocks the Theory That Instinct's Valuation Has Soared

    Lukas Gruber said Instinct is raising funds at a $30 billion valuation, with its valuation climbing from $100 million all the way to about $30 billion within a year, and believes wealth creation will tilt toward private investors. Peter Yang used this to mock the idea that, extrapolating at this rate, its market cap would be at least $3 trillion in ten years, and said Instinct is a good product, but this tweet doesn't hold up.

    Quoted postLukas Gruber · @lukasgruber__

    Instinct is currently raising @ $30b. $100M → $500M → $2.5B → $10B → ~$30B in sub 1y. If you're still not inside the private market this should be a good signal to pivot aggressively. Wealth creation is going to be heavily skewed toward private investors over the coming decade.

    #English text is being prepared.#Industry news

  18. Electronic Technology Application
    AI score63

    Oracle Expands NVIDIA Partnership and Adds Google TPU and Cerebras Chips

    Oracle announced that it will deploy NVIDIA's latest-generation Vera Rubin AI servers at scale, reaching several gigawatts, while also bringing in Google TPU and Cerebras chips to advance supply diversification. Oracle will become the first hyperscale cloud provider other than Google to offer Google TPU; the service is expected to launch in 2027 and support large-model training, inference, and agent applications. Oracle also plans to offer SambaNova chips and to switch to its in-house Acceleron networking chips in systems based on NVIDIA chips.

    Same story · 9 reports

    #Industry news#English text is being prepared.#Google

  19. Huxiu
    AI score36

    Civil service exam applicant numbers hit a new high, so why are civil service exam prep providers like Huatu Shanding and Fenbi finding it harder and harder to make money?

    In the 2027 national civil service exam, 3.025 million people passed qualification review to compete for 35,691 slots, yet Huatu Shanding's revenue fell 10.49% and net profit fell 84.71% in the first half of 2026, while Fenbi's revenue dropped about 16.3% and it swung from profit to loss. The contraction in recruitment scale directly squeezed the high-priced interview training market; Huatu Shanding said that the 2026 national and provincial exams together recruited about 216,700 people, down 14.3% year on year. Free question banks, low-priced courses and AI grading have weakened pricing for foundational courses; Fenbi's training revenue fell 17.5% and its gross margin dropped from 54.0% to 41.8%.

    #Industry news#English text is being prepared.

  20. HuxiuOpinion
    AI score57

    Why Tesla and SpaceX Still Want to Merge: The Capital and Shareholder Challenges Behind a $3.5 Trillion Deal

    Musk publicly supports a Tesla-SpaceX merger but acknowledges a conflict of interest and will not directly participate in negotiations between the two sides' special committees. Both companies have separately established independent director special committees to study a potential transaction. Based on market valuations at the time, SpaceX was worth about $2 trillion and Tesla about $1.5 trillion, for a combined total of roughly $3.5 trillion. The deal is still in the research stage and no terms have been reached. The article notes that the two sides already collaborate on the Terafab chip project, AI compute, and Megapack energy storage procurement. The core of the merger lies in unifying capital allocation, but the exchange ratio involves the distribution of hundreds of billions of dollars in shareholder interests, and it must also contend with the SolarCity litigation precedent and cross-border regulatory review.

    Same story · 2 reports

    #Industry news#English text is being prepared.#English text is being prepared.