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.
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.
Chinese AI startup DeepSeek has released its latest open-source large language model DeepSeek V2, featuring a 236 billion parameter Mixture of Experts (MoE) architecture with inference costs only 1/30th of OpenAI's GPT-4o, rapidly topping Hugging Face's trending charts with over 80,000 interactions.
Alibaba Cloud's newly released Qwen2.5-Max model has achieved 86.1% on the authoritative Chinese MMLU benchmark, surpassing OpenAI's GPT-4o (85.8%) to claim the top spot among Chinese language models. This breakthrough has rapidly gained traction in the open-source community, with Hugging Face downloads surging by over 100,000 within 24 hours.
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.
DeepSeek team released V2 open-source model, excelling in Chinese benchmarks especially in math and code generation, surpassing Meta's Llama3 while costing only 1/10 to train. The breakthrough ignited discussions on X platform's Chinese community with over 500,000 views.
Chinese AI startup DeepSeek has released DeepSeek-V2, a large language model that outperforms GPT-4o in Chinese math reasoning benchmarks while achieving high efficiency through its MoE architecture with 236B total parameters but only 21B active.
