Google's announcement of AI-driven search features has raised expectations, but the lack of details reveals deeper challenges. The announcement reflects Google's strategic responses to industry pressures and internal hurdles.
Chinese AI startup DeepSeek has released its latest large language model DeepSeek-V2, featuring 236B parameters with MoE architecture, surpassing Claude 3.5 Sonnet in Chinese math reasoning and code generation, marking a significant breakthrough for open-source models.
Chinese AI startup DeepSeek has released DeepSeek-V2, a 236B parameter MoE model that matches GPT-4o's performance while costing only 1/30 for inference. The release has sparked widespread discussion with over 10,000 GitHub stars and 150,000+ mentions on X platform.
NVIDIA Research has open-sourced PersonaPlex-7B, a 7-billion parameter AI model designed for real-time voice interaction with full-duplex capability. The model enables natural interruptions and simultaneous listening and speaking, marking a significant advancement in conversational AI.
Beijing-based Moonshot AI officially launched Kimi k1.5, a large language model supporting up to 2 million characters of context length, setting a new record in long-text processing capabilities. As a free and open domestic AI product optimized for Chinese language, Kimi k1.5 has quickly sparked bilingual discussions on X platform.
Meta officially launched Llama 3.1 series on July 24th Beijing time, with the 405B parameter version achieving 88.6% on MMLU benchmark, crowning it as the pinnacle of open-source large language model performance. The model not only excels in multilingual support and long-context processing but also offers free commercial licensing as a fully open-source solution, quickly igniting enthusiasm in the developer community.
Chinese AI startup DeepSeek launched its V2 chatbot based on a new large language model, garnering over 100,000 discussions on X within a day. Users praise its strong performance in programming and mathematics, with Chinese language capabilities matching OpenAI's GPT-4o, all while being completely free.
Alibaba Cloud officially launched the Tongyi Qianwen Qwen2.5-Max large language model, which outperforms Google's Gemini 1.5 Pro in mathematics and coding benchmarks. The open-source free strategy quickly went viral in the Chinese AI community with over 30,000 reposts.
Meta AI has released the Llama 3.1 series of large language models, with the 405B parameter flagship version available as fully open-source. This move instantly ignited enthusiasm in the open-source community, with model downloads surpassing one million on the first day and related discussions on X platform exceeding 300,000 posts.
Recently, an error log from Google Vertex AI leaked potential details about Anthropic's next-generation AI model Claude Sonnet 5, codenamed "Fennec," which reportedly outperforms Claude Opus 4.5 across all metrics while offering more competitive pricing.
Meta AI officially released Llama 3.1 series open-source large language models on July 24, with the flagship 405B parameter version delivering stunning performance that refreshes industry perceptions. The model performs on par with top closed-source models like GPT-4o and Claude 3.5 Sonnet across multiple authoritative benchmarks, while supporting a longer 128K context window and multilingual capabilities, marking the entry of open-source AI into the "giant beast" era.
