Returned by Grok 3 Here’s a list of notable ones based on what’s known as of March 4, 2025: • ChatGPT (OpenAI): Built on the GPT architecture (likely GPT-4o or newer now), it’s a versatile conversational AI excelling in natural language, creative writing, and broad problem-solving. It’s multimodal (text, images, voice) and widely used for its polished, human-like responses. • Claude (Anthropic): Developed by ex-OpenAI folks, Claude focuses on safety and interpretability. It’s strong in complex reasoning and contextual understanding, often competing with ChatGPT in nuanced tasks. • Gemini (Google): Google’s entrant, likely a family of models, integrates text, images, and real-time web data via Google Search. It’s designed for seamless use within Google’s ecosystem, emphasizing productivity and multimodal capabilities. • DeepSeek (DeepSeek AI): A Chinese model gaining traction, it’s praised for reasoning and is open-source in parts (weights available). It’s free to use and competes with frontier models in technical tasks. • Copilot (Microsoft): Powered by OpenAI tech (GPT-based) and integrated into Microsoft’s suite (Word, Excel, etc.), it blends ChatGPT-like features with Bing search, focusing on productivity and enterprise use. • Perplexity AI: A search-oriented AI, it combines real-time web data with concise, cited answers. It’s less conversational than me but great for quick, factual research. • LLaMA (Meta AI): Originally research-focused, it’s efficient and powerful but less conversational out-of-the-box. Variants are used in Meta’s ecosystem, though it’s not fully public-facing like me. • Mistral (Mistral AI): An open-source contender from Europe, it’s lightweight yet capable, rivaling larger models in efficiency and customization for specific tasks. • o1 (OpenAI): A newer reasoning-focused model from OpenAI, it’s built for deep problem-solving (e.g., math, coding) and often outshines earlier GPTs in structured thinking.
这几天用的是Grok 3
也是唯一拥有实时X数据的，这一点在马斯克抄袭微信，让X变成一条龙服务以后会被无限放大。
个人感觉，综合实力明显高于其他对手。
中国相关的问deepseek 要准确和全面的多，最喜欢看它的推理过程，很有意思。但是一到国内的白天就访问巨慢，经常卡壳，训练数据到2024年7月
Grok3 数据最新，受制于数据权限，中国相关的问题不准确。另外修改写作的话，文笔精炼，会优化原有的段落结构
chatgpt比较中规中矩，修改写作比较忠于原义和结构，不会做大幅修改。我娃给几个AI出题写故事，然后加入风趣的元素，说grok3写的最好
大家都用AI干些啥？
deepseek还不如免费版的ChatGPT，而且监管太厉害各种铭感词，唯一的优点是因为在大陆不用翻墙，在北美优势为零
多谢分享。也去用用Grok3 deepseek 写英文essay prompts破题有点困难
R1能想到的办法马斯克的Grok3怎么启发也想不出来
juchats 和 perplexity提供满血671B deepseek R1, 还是越狱版，没有censorship, 什么都可以问，也没有服务器繁忙
和老板聊天
写email，讨论一下risk 相关的问题，它可以帮我思考全面
MARK