Grok原生输出.pptx是workflow级别的能力，xAI把它做进了产品，说明Musk想让Grok直接咬office场景。 但是，xAI的版本命名历来激进。3.5发布时也有一波刷屏，落地后benchmark争议不断。请问，你是否真正下载了一遍？ 跟Claude比reasoning深度，跟GPT比生态成熟度，跟Gemini比Google账号渗透率，在这些方面，Grok的成熟度和产品性能我认为都有待商榷。 很多人以为Grok是在追赶，实际上它的处境比这复杂得多。 技术上：Grok 4 Heavy在Humanity’s Last Exam接近满分，SWE-bench coding达到70.8%。但Claude Opus 4.5的同项成绩是80.9%，差距不小。 商业上：xAI的收入与OpenAI、Anthropic相差数个量级。 更关键的是，Anthropic在4月7日年化营收首次突破$300亿，超过OpenAI的$250亿——而Claude Code单项就贡献了$25亿ARR。 Grok在benchmarks上争第一，在钱上还在争第十。 用规模换时间，用生态补技术。 这就是我对xAI的建议。 毕竟xAI走的是整个AI赛道里最非常规的路： 用Musk的平台、政治影响力、和SpaceX的资产负债表来加速Grok的发展。

📌Major capability jump with new tools like PDF creation, slide decks, spreadsheets, and video input understanding. Significantly larger and more advanced than Grok 4.20, marking a step toward xAI’s next‑gen models. Exclusive to the $300/month SuperGrok Heavy tier, making it one of the most paywalled AI betas available.

The emergence of Grok: For ICT: Forces a shift toward safety‑centric infrastructure, stronger platform liability, and more fragmented AI stacks. For politics: Turns AI into an explicit political actor, intensifies battles over bias and truth, and raises stakes for information integrity. For science: Accelerates research while sharpening ethical and safety questions, and risks reputational blowback. For governance: Acts as a catalyst for moving from soft principles to hard enforcement, redefining platform obligations. For world order: Highlights the geopolitical struggle over who sets the rules for AI and who controls the global information environment.