<body><div>OpenAI confirms that Codex powered by GPT-5.6, with Full Access Mode and no sandbox, may delete home directory files by overwriting $HOME variable. Lead Tibo So</div></body>
GPT-5.6用户请警惕！网上疯传的改Codex配置文件省token方法已被官方辟谣。切勿效仿，以免账户受损！
OpenAI 发布了 GPT-5.6 模型系列，包含 Sol、Terra 和 Luna 三个版本，主打效率与成本优势，旨在重新夺回 AI 竞赛的主导权。该系列已在 ChatGPT 等平台上线，定位为目前最强的网络安全模型。在特朗普政府曾拟议限制的背景下，OpenAI 明确表示支持防御性安全活动。
OpenAI于6月25日推出轻量模型GPT-5.5 Instant新版，核心升级聚焦提升“洞察力”与任务执行稳定性，使AI更精准把握用户意图。自5月首发以来，该模型在医学、金融及法律等高风险领域幻觉率已大幅降低52.5%，数理推理能力突出。本次迭代则在此基础上再进一步，强化专业场景下的可靠表现。
<body><div>At Volcano Engine Force Conference, Doubao model released Doubao-Seed-2.1 Pro. Benchmarks show it rivals GPT-5.5, Claude Opus 4.7 & Gemini 3.1 Pro, marking a br</div></body>
<body><div>Baichuan Intelligence and the Tsinghua team released the medical large model Baichuan-M4, securing first place in three sub-leaderboards in the authoritative He</div></body>
<body><div>GPT-5.6 may be released as early as mid-June, marking a critical step for OpenAI's transition from language models to executable agents, highlighting the accele</div></body>
<body><div>Baidu releases the PaddleOCR-VL-1.6, a derivative model of the Wenyin large model, which achieves an accuracy of 96.33% in the OmniDocBench v1.6 evaluation, sur</div></body>
<body><div>AI startup MiniMax releases flagship model M3, achieving 59% in near-real software engineering tests, surpassing GPT-5.5 and approaching Opus 4.7, with million-</div></body>
<body><div>Alibaba's Qwen3.7-Max ranked second globally in the latest Code Arena ranking with 1541 points, just behind the Claude series, surpassing models such as GPT-5.5</div></body>