Large Model

Xiaomi's open-sourced 7 billion parameter inference macromodel, which outperforms models such as OpenAI o1-mini in mathematical reasoning and code competitions by a small margin.

Ema introduces a hybrid expert modeling system that dynamically combines multiple models to accomplish enterprise-class AI tasks at low cost and high accuracy.

Baidu has launched a new generation of high-level AI assistants to disassemble complex tasks and automate the entire process with autonomous deep thinking, multimodal toolchain invocation and extreme cost advantages.

Baidu launched a multimodal strong inference AI model, the cost of which is directly reduced by 80%, supports cross-modal interaction and closed-loop invocation of tools, and empowers enterprises to innovate intelligently.

Amazon has introduced a new generation of generative AI speech models with unified model architecture, natural and smooth voice interaction, real-time two-way conversation capability and multi-language support, which can be widely used in multi-industry scenarios.

The world's largest biology AI model, jointly developed by multiple top organizations, is trained based on massive genetic data and can accurately predict genetic variants and generated sequences to help breakthroughs in life sciences.

The world's first big model of music reasoning introduced with thought chain technology released by KunlunWanwei supports multi-style and emotional music generation, song reference and tone cloning with low latency and high quality performance, and opens up API services for enterprises and developers to integrate the application.

Google introduces advanced AI models with powerful reasoning capabilities, multimodal support, and ultra-long context windows for multiple scenarios such as academic research, software development, creative work, and enterprise applications.

Tencent's self-developed deep thinking models with fast response, ultra-long text processing and strong reasoning capabilities have been widely used in intelligent Q&A, document processing and other fields.

Cohere released a lightweight AI model with powerful features such as efficient processing, long context support, multi-language and enterprise-grade security, designed for small and medium-sized businesses to achieve superior performance with low-cost hardware.

High-performance inference models from OpenAI with enhanced multimodal inference capabilities, structured outputs, and function call support, designed to handle complex professional problems with high pricing but high performance.

Anthropic's top-of-the-line AI models for hardcore developers tackle ultra-complex tasks with powerful code processing and a 200k context window.

Baidu launched an advanced large language model with deep thinking, multi-modal support and multi-tool invocation capabilities to meet the needs of multiple domains with excellent performance, affordable price and rich functionality.

Baidu's self-developed native multimodal basic big model, with excellent multimodal understanding, text generation and logical reasoning capabilities, using a number of advanced technologies, the cost is only 1% of GPT4.5, and plans to be fully open source.

The Tsinghua University team and Qingcheng Jizhi jointly launched an open source large model inference engine, aiming to realize efficient model inference across chip architectures through underlying technological innovations and promote the widespread application of AI technology.

Byte Jump launched a native bilingual image generation model with excellent comprehension and rendering capabilities for a wide range of creative design scenarios.

Google launched a new generation of open source AI models with multi-modal, multi-language support and high efficiency and portability, capable of running on a single GPU/TPU for a wide range of application scenarios.

Alibaba's open-source multimodal large language model uses RLVR technology to achieve emotion recognition and provide an interpretable reasoning process for multiple scenarios.

Alibaba released a high-performance inference model with 32 billion parameters that excels in mathematics and programming for a wide range of application scenarios.

OpenAI's large-scale language model, officially launched on February 28, 2025, is an upgraded version of GPT-4.

Alibaba launched an efficient video generation model that can accurately simulate complex scenes and actions, support Chinese and English special effects, and lead a new era of AI video creation.

Anthropic has released the world's first hybrid reasoning model that demonstrates superior performance and flexibility by being able to flexibly switch between rapid response and deeper reflection based on different needs.

Alibaba's open source multimodal large language model with powerful visual understanding, OCR, video processing and reasoning capabilities, supporting multiple scale versions.

The third generation of artificial intelligence models developed by Musk's xAI company, with superior computational and reasoning capabilities, can be applied to a variety of fields such as 3D model generation and game production, which is an important innovation in the field of AI.

NetEaseYouDao launched the first 14B lightweight model in China that supports step-by-step reasoning and explanation, designed for educational scenarios, which can help students efficiently understand complex math problems.

Google DeepMind launches a 100 billion visual language dataset designed to enhance the cultural diversity and multilingualism of AI models.

Developed by the DeepSeek team, it is an efficient visual language model based on a hybrid expert architecture with powerful multimodal understanding and processing capabilities.

Hangzhou Depth Seeker has launched an efficient open source language model with 67.1 billion parameters, using a hybrid expert architecture that excels at handling math, coding and multilingual tasks.

An AI model developed by Fei-Fei Li's team that achieves superior inference performance at a very low training cost.

Google released a high-performance AI model with strong coding performance and the ability to handle complex cues with a contextual window of 2 million tokens.