AI Agent News
实时追踪 AI Agent 赛道的重大事件、融资动向、模型发布和技术突破
最新行业资讯
实时追踪 AI Agent 赛道的重大事件、融资动向、模型发布和技术突破
重大事件时间线
OpenClaw GitHub 爆发
OpenClaw 10 天冲上 GitHub 全球 Top 10，超越 Linux 内核 Star 增速
Meta 20亿收购 Manus
Meta 以 20 亿美元收购 Manus AI，通用 Agent 赛道正式被巨头锁定
DeepSeek-V3 开源
性价比之王，成本仅 GPT-4 的 5%
Manus 一夜爆火
全球首款通用 AI Agent 在国内社交平台引发空前关注
OpenAI Deep Research
OpenAI 推出深度研究 Agent，一键生成专业研究报告
MCP Server 破 500
MCP 生态爆发，3 个月构建 500+ Server
DeepSeek-R1 震惊全球
开源推理模型，成本仅 OpenAI 的 3%，引发全球 AI 格局震动
MCP 协议诞生
Anthropic 发布 Model Context Protocol，成为 Agent 接口事实标准
Claude Computer Use
Anthropic 让 AI 首次直接操控电脑屏幕，开创计算机使用新范式
Replit Agent 全栈自动化
自然语言到上线产品，面向非工程师
Cursor ARR 破亿
史上增长最快 SaaS，AI 编程工具新王者
Claude 3.5 登顶 SWE-bench
最强编程 AI，Bug 修复能力达到初级工程师水平
Devin 发布
全球首个自主 AI 软件工程师，能独立完成完整编程任务
OpenAI完成400亿美元融资，估值达3000亿：AI独角兽迈入超级巨头时代
OpenAI宣布完成400亿美元新一轮融资，估值达3000亿美元，成为全球估值最高的私营科技公司。软银以150亿美元领投，微软追加100亿美元。融资将用于扩大算力基础设施、AGI研究和全球市场拓展。OpenAI收入预计2025年突破120亿美元，同比增长300%。
Figure 02机器人工厂部署：OpenAI加持的人形机器人开始正式"上班"
Figure AI宣布其Figure 02人形机器人已在BMW斯帕坦堡工厂正式投产，执行汽车零件组装任务。机器人使用OpenAI多模态模型理解工厂指令，通过视觉语言行动模型执行操作。初期效率为人类工人的30%，计划一年内提升至70%。Figure CEO表示2025年将部署100台机器人。
摩根大通AI交易系统每日处理万亿美元：AI如何重塑华尔街
摩根大通披露其AI驱动的交易系统已处理所有股权交易的70%，固收交易的40%。AI系统负责最优执行路径选择、流动性预测和交易成本分析。AI风控系统每天识别并拦截超过40亿美元的可疑交易。摩根大通AI研发投入已超过20亿美元/年。
Bloomberg发布AI金融分析平台：GPT-4驱动的财务数据解读
Bloomberg宣布推出Bloomberg AI，整合GPT-4技术提供自然语言财务数据查询、财报解读和市场分析能力。专业用户可以用自然语言直接查询Bloomberg终端数据，生成定制化分析报告。Bloomberg AI已在全球5000家金融机构开始试用。
Cognition AI Devin 2.0: Autonomous Software Engineering Hits Production Scale
Cognition AI released Devin 2.0, its autonomous software engineering agent, with significant capability improvements. The new version handles full feature development cycles—from reading requirements to writing code, running tests, fixing bugs, and opening PRs. Enterprise pilots at 50+ companies report Devin completing 30-40% of routine engineering tasks autonomously. New capabilities include codebase indexing for context (up to 1M lines), multi-file refactoring, and integration with Jira, Slack, and GitHub Actions. Pricing: $500/month per seat with enterprise volume discounts.
OpenAI Releases GPT-5: Major Leap in Reasoning and Multimodal Capabilities
OpenAI announced GPT-5, its most capable model, achieving new state-of-the-art scores on MMLU (92%), HumanEval (97%), and MATH (90%). The model features native multimodal input (text, image, audio, video), 1M token context window, dramatically improved reasoning with reduced hallucination rate, and real-time web browsing. Enterprise customers report 40-60% improvement in complex task completion vs. GPT-4o. Available via API at $15/1M input tokens, $60/1M output tokens, with enterprise volume discounts.
Anthropic Releases Claude 3.7: Best Coding and Scientific Reasoning Model
Anthropic released Claude 3.7, achieving new state-of-the-art on coding benchmarks (HumanEval: 95.2%, SWE-Bench: 49.5%) and scientific reasoning. The model features an improved "extended thinking" mode for complex multi-step problems, 200K context maintained with improved mid-context accuracy, and new computer use capabilities for browser automation. Anthropic's model card transparency reports show Claude 3.7 has the lowest hallucination rate in independent evaluations. Available on Claude.ai and API at same pricing as Claude 3.5.
Apple Intelligence iOS 19新特性：端侧AI全面升级，隐私保护再加强
Apple在WWDC 2025宣布iOS 19 Apple Intelligence重大升级，新增端侧推理能力（支持30B参数模型）、跨App AI自动化、增强版Siri（支持多轮对话和屏幕内容理解）。私有云计算架构确保用户数据不离设备，差分隐私技术保护个性化学习数据。
OpenAI Launches Codex Cloud: Autonomous Coding Agents in the Cloud
OpenAI announced Codex Cloud, a cloud-based environment where AI coding agents run autonomously on software tasks. Users assign tasks via natural language; Codex reads code, writes implementations, runs test suites, and iterates until tests pass—all in isolated cloud environments. Multiple parallel agents can work simultaneously on different tasks. Beta users report completing full features in 1-3 hours that would take developers half a day. Codex Cloud integrates with GitHub, creating PRs automatically. Pricing: usage-based at $0.04-0.15 per task depending on complexity.