Search AI Products and News

Large Language Models (LLMs) have achieved significant progress in complex reasoning tasks by combining task prompts with large-scale reinforcement learning (RL), as demonstrated by models like Deepseek-R1-Zero, which directly apply reinforcement learning to base models, showcasing strong reasoning capabilities. However, this success is difficult to replicate across different base model families, especially within the Llama series. This raises a core question: what factors lead to inconsistent performance of different base models during reinforcement learning? How does reinforcement learning perform in

The plan's advancement has triggered dissatisfaction among developers. According to VentureBeat, OpenAI's current API deprecation involves the GPT-4.5 Preview version, which is an intermediary model launched between GPT-4 and GPT-5. Multiple developers have stated that this sudden change has disrupted their development plans, forcing significant adjustments to applications already built based on this API. A developer who wished to remain anonymous commented: We understand the necessity of technological iteration, but the lack of sufficient buffer time in the schedule has brought additional pressure to development teams.

At the recently held 'Force Primordial Power Conference', ByteDance’s Volcano Engine officially launched the DouBao large model 1.6 and video generation model Seedance 1.0pro. This release marks another leap forward in AI technology, particularly in multi-modal understanding and video generation, demonstrating strong market competitiveness. DouBao 1.6 has performed excellently in multiple international authoritative evaluations, especially excelling in complex reasoning and competition-level mathematics tests. Tan Dai, President of Volcano Engine, stated that this model not only supports multi-

Recently, the DeepSeek team released a technical paper about their latest model, DeepSeek-V3, focusing on the challenges of scaling in large AI model training and related considerations of hardware architecture. This 14-page paper not only summarizes the experiences and lessons learned during the development of V3 by DeepSeek but also provides deep insights for future hardware design. Notably, CEO of DeepSeek, Liang Wenfeng, also participated in writing the paper.

The Gemini API has launched an implicit caching feature, offering developers significant cost optimization experiences. This new functionality eliminates the need for developers to manually create caches; when requests share a common prefix with previous requests, the system automatically triggers cache hits, providing up to a 75% discount on tokens. This update now covers the Gemini2.5Pro and 2.5Flash models, further enhancing the cost-effectiveness of AI development. For more details, please refer to the official website: https://develo