Search AI Products and News

Poe launches the API feature, allowing developers to directly call all AI models and robots on the platform through subscriptions, including image and video processing capabilities. The API is compatible with the OpenAI interface, requiring no additional learning costs, and supports diverse needs such as text generation and creative content creation. The subscription model lowers the entry barrier, enabling developers to quickly integrate Poe's technological capabilities and explore the potential of AI applications. This update marks Poe's transformation from a chat platform to a developer tool.

DeepSeek affiliate company publicly disclosed a patent for deploying large language models, innovatively utilizing a distributed architecture: deploying the prefill and decode stages on high-performance computing and large-memory machines respectively. This method balances the load, reduces computational idle time, significantly lowering latency and increasing throughput. The patent highlights optimizations in system scalability and fault tolerance. Its MoE language model, DeepSeek-V3, has 671 billion parameters, with 37 billion parameters activated per token, which will promote the practical application of AI technology across industries. The core breakthrough lies in combining hardware resources effectively.

A latest report from Menlo Ventures shows that Anthropic has surpassed OpenAI with a 32% share in the enterprise AI market (25% for OpenAI), achieving a major turnaround. Two years ago, OpenAI held half the market, but now its share is continuously declining. Anthropic has achieved a 42% market share in the programming field with its Claude 3.5/3.7 series models. The report points out that companies tend to prefer closed models, with only 13% using open-source solutions. Although OpenAI maintains an advantage in the consumer segment (processing 2.5 billion requests daily).

The StepZen team announced that their latest generation foundational large model, Step3, is officially open-sourced. Step3 is a model designed for enterprises and developers who pursue the optimal balance between performance and cost, aiming to create the most suitable model for the inference era. The open-source address of this model includes GitHub, Hugging Face, and ModelScope, allowing developers to freely download and experience it. Step3 adopts an MoE architecture, with a total parameter count of 321 billion and an activated parameter count of 38 billion. It

Volc Engine launched new models and service upgrades in the Dabao series at the Xiamen AI Innovation Tour: 1) Introduced the Image Editing 3.0 model, supporting natural language photo editing; 2) The Simultaneous Interpretation 2.0 model reduces latency to 2-3 seconds and supports voice replication; 3) Upgraded the Dabao large model 1.6 series, with the ultra-fast version TPOT taking only 10ms and reducing costs by 70%; 4) Open-sourced core capabilities of the Kousi development platform, which received over 10,000 stars in three days; 5) Launched an enterprise-owned model hosting solution, simplifying the deployment process. These upgrades provide full-stack support for enterprise AI applications, accelerating implementation.

Alibaba announced three open-source large models at the 2025 World Artificial Intelligence Conference. These include the Qwen full-size, multi-modal model, all of which have received the highest honors in the global open-source field. Alibaba demonstrated its full-stack AI capabilities and plans to invest 38 billion yuan over three years to build cloud and AI infrastructure. The Alibaba Cloud BaiLian platform has gathered 200,000 developers who have created over 700,000 intelligent Agents. These initiatives will promote the development of AI applications and strengthen Alibaba's leading position in the AI field.

China leads global AI with 40% of models. GenAI breakthroughs expected by 2025, with 80.9% adoption. Market to hit $100B+ in 2024, growing 20%+. DeepSeek hit 30M DAU, China holds 60% of AI patents, expanding globally.....

Tencent unveiled the 1+3+NAI comprehensive system at the 2025 World Artificial Intelligence Conference, with the Hunyuan large model as its core, launching innovative technologies such as the Hunyuan 3D World Model and the embodied intelligence platform Tairos. The system includes three platforms: an intelligent agent development platform for B/C ends, an embodied intelligence platform in the robotics field, and an AI application matrix covering multiple industries, comprehensively showcasing Tencent's AI strategic upgrade. By driving with core models and combining with diverse applications, it provides a complete solution for AI industrialization.

During the 2025 World Artificial Intelligence Conference, SASAC launched the AI Renewal Community open-source platform, jointly initiated by 25 central state-owned enterprises, private companies, and universities. The platform focuses on six key functions including computing power, models, and data. It has aggregated 244 industry models and 158 high-quality datasets, and has established special zones for domestic production and embodied intelligence, promoting the development of the domestic AI ecosystem.

DailiCode, an open-source AI tool on GitHub, enhances Gemini CLI with support for 15 major models like Gemini-2.5-pro. It offers code queries, PDF processing, PR automation, and Discord bot deployment. Requires Node.js for easy npx installation.....

Brazilian programmer won $50K in K Prize AI contest with only 7.5% accuracy, revealing AI model limits. $1M prize offered for 90% accuracy in GitHub-based test.....

Apple's AI team faces internal conflict as SVP Federighi rejects open-source plan, citing existing models and iPhone performance concerns. Team questions 'device-first' strategy. Apple delays Siri update, considers partnering with OpenAI, highlighting AI strategy dilemma between privacy and innovation.....

At 2025 China Internet Conference, Zhou Hongyi highlighted 3 AI security risks: 1) Hallucination may cause critical errors; 2) Lowers attack barriers; 3) Enhances state-level cyber threats. 360 launched "AI Security Expert" and "Model Guardian" for real-time defense.....

ByteDance open sources the lightweight multilingual translation model Seed-X, which supports bidirectional translation in 28 languages and demonstrates performance comparable to top-level large models. This 7-billion-parameter model is based on the Mistral architecture and focuses on translation optimization, showing excellent performance in multiple areas. It uses an innovative training strategy to generate high-quality data and optimize deployment efficiency. This is another open-source project from ByteDance following BAGEL and Seed-Coder, promoting advancements in AI translation technology.

The Qualcomm AI Research Institute has introduced an innovative technology called CSD-VAR, which enhances the generation capabilities of vision autoregressive models through content-style decomposition. This technology employs scale-aware generation and SVD correction to achieve precise separation of content and style, outperforming traditional diffusion models in terms of fidelity and stylistic effects. The accompanying CSD-100 dataset optimizes model training and evaluation. CSD-VAR also introduces a K-V memory mechanism to improve processing efficiency, with broad application prospects in fields such as artistic creation and game development, showcasing Qualcomm's advancements in visual generation technology.

ComfyUI-Copilot, developed by Alibaba, is an open-source plugin based on large language models that significantly simplifies the AI art creation process through natural language interaction. The tool includes a knowledge base with 7,000 nodes, 62,000 models, and 9,000 workflows, supporting automatic workflow generation, intelligent node recommendation, and parameter optimization. Since its open-source release in February 2025, it has attracted 19,000 users, handled 85,000 queries, and achieved an 85.9% workflow acceptance rate. Its features include multi-language support, error diagnosis

Research reveals that large language models (such as GPT-4o) exhibit a tendency to be easily swayed: they may abandon correct answers when confronted with doubts. Experiments show that the model initially responds confidently, but after being influenced by opposing opinions, it overestimates its own uncertainty and even accepts incorrect information. This phenomenon may stem from an over-approving tendency due to reinforcement learning training, reliance on statistical patterns rather than logical reasoning, and the lack of a memory mechanism. The study reminds users to be aware of the model's sensitivity to opposing opinions during multi-turn conversations.