| Mobile Web

Search results for ByteDance

AI & Enterprise

Big Tech steps up coding AI push, arms with China AI performance

Big Tech companies are accelerating an offensive in the coding AI tools market, raising questions about shifts in a field led by Anthropic, OpenAI and Cursor. Meta unveiled its first AI coding agent, Muse Code, in beta. Google is in talks on a deal worth about $1.5 billion involving AI coding startup Mechanize. AWS launched Kiro Crew, an autonomous workspace. Chinese firms also quickened moves to build large-scale AI models.

AI & Enterprise

ByteDance training AI model with up to 10 trillion parameters, report says

ByteDance is training an AI model with up to 10 trillion parameters, which could be close in scale to Anthropic’s latest Mythos system, Reuters reported on Aug. 7, citing the Financial Times. The FT cited sources familiar with the matter. The report said 10 trillion parameters would be more than three times larger than Moonshot AI’s Kimi K3. The ByteDance model is in a pretraining stage that typically takes 3 to 6 months, and is expected to be released after fine-tuning.

AI & Enterprise

ByteDance founder Zhang Yiming rules out distillation for AI model training

ByteDance founder Zhang Yiming (장이밍) has made clear the company will not use “distillation” techniques to train its AI models, even if it costs more. He has recently become directly involved in the company’s AI strategy, reviewing development progress and hiring. The stance comes as the United States steps up scrutiny of Chinese firms’ use of distillation. ByteDance’s approach is tied to development of its video generation model, Seedance, and large investments in GPUs, data and talent.