Anthropic Unveils Claude Mythos, OpenAI's GPT-6 Nears Release, and Tencent Launches AI Browser

Anthropic has announced Claude Mythos Preview, a new artificial intelligence model described by the company as a significant leap in performance. Boris Cherny, a co-creator of the CC programming language, characterized Mythos as "incredibly powerful." Anthropic stated it does not plan to make Claude Mythos Preview publicly available yet, aiming instead to develop safety guardrails before broad deployment. The company intends to test new safety technology on an upcoming model, Claude Opus.

Digital network with glowing protective shields, symbolizing AI safety guardrails.

Concurrently, Anthropic introduced Project Glasswing, an AI cybersecurity initiative built on the Mythos model. Initial partners include Amazon Web Services, Apple, Cisco, and Google. Over 40 additional organizations involved in critical software infrastructure have also been granted access.

Abstract digital shields and data streams symbolizing AI cybersecurity.

AI Model Developments

OpenAI's next-generation model, codenamed "Spud," has completed pre-training and is anticipated for release around April 14, according to tech blogger "iruletheworldmo" on the X platform. OpenAI CEO Sam Altman previously indicated in mid-March that the company was training "the world's best model" at its Abilene, Texas, facility, a statement believed to refer to GPT-6. This development follows controversies for OpenAI, including the cessation of its Sora video generation service on March 24 and recent reports of Elon Musk's efforts to remove Altman and President Greg Brockman from their positions.

Modern server room with glowing racks, symbolizing AI model pre-training.

DeepSeek launched an "expert mode" for its AI on April 8, renaming its previous offering to "fast mode." The fast mode is designed for daily conversations and quick responses, while the expert mode is optimized for complex problems, though it may experience wait times during peak usage. The expert mode currently does not support file uploads or multimodal functions.

Zhipu Qingyan integrated its new flagship model, GLM-5.1, on April 8. The company reported that GLM-5.1 is optimized for long-horizon tasks, supporting continuous multi-round autonomous execution for up to eight hours. This capability allows it to complete engineering-grade deliverables such as code writing, solution design, and system construction. Zhipu Qingyan stated that GLM-5.1 achieved a new global record in the SWE-bench Pro benchmark test, surpassing GPT-5.4 and Claude Opus 4.6.

Winding digital pathway with glowing nodes, symbolizing long-horizon AI tasks.

Aishen Technology officially released PixVerse C1 on April 8, which it described as the world's first large model for the film and television industry. PixVerse C1 is a vertical model trained for creating short dramas, animations, and other film and television content. It supports text-to-video, image-to-video, reference-based generation, and first/last frame capabilities. The model can generate 1080P videos up to 15 seconds long with synchronized audio and video, and automatically segments scenes based on prompt commands.

Film editing suite with multiple screens showing video production stages, symbolizing AI in film.

Tencent and Google AI Initiatives

Tencent launched its first browser, "Lobster" (QBotClaw), on April 8. Users can access it via the "AI" button in the QQ browser on Mac devices, with a Windows version planned for future release. QBotClaw features a self-developed QQ browser Skill, deep memory capabilities, and direct integration with WeChat.

Abstract glowing digital heart with protective data streams, symbolizing AI mental health support.

Google announced several initiatives on April 7 aimed at enhancing mental health support through artificial intelligence. The Gemini model has been updated with a "help module" that, upon detecting potential mental health crisis signals in user conversations, offers a one-click option to connect to professional mental health hotlines. Google.org, the company's philanthropic arm, committed $30 million to fund mental health hotline services globally and to use AI simulation platforms for volunteer training. Additionally, Google plans to strengthen emotional isolation protections for minors, ensuring AI assists rather than replaces professional diagnosis and treatment.

Human silhouette against a vast, glowing AI network, symbolizing human scarcity in the AI era.

Separately, at the AI FUT Conference in Beijing on April 8, Luo Zhenyu discussed the concept of "human scarcity" in the AI era. He suggested that "telling the world what it wants is the last scarcity and value of humans," and emphasized that individuals should focus on becoming unique rather than relying on AI for everything.