匈牙利央行：2025年下半年，家庭贷款显著增长。房地产市场风险已增加。

Third-party firm Artificial Analysis’s latest global large-model ranking gave Qwen3.7-Max a score of 57, close to top GPT, Claude and Gemini models, and ranked it first among Chinese models. The model handles programming tasks from page prototyping to complex multi-file projects; in office workflows it can integrate tools and coordinate multi-agent collaboration to automate multi-step processes. In fully autonomous tests it remained coherent over runs exceeding 35 hours and more than 1,000 tool calls. It also demonstrates cross-agent-framework generalization, performing well when connected to various agent frameworks and tools.

Anthropic is granting 150 additional institutions worldwide access to Mythos, its AI model for identifying cybersecurity vulnerabilities, bringing total access to about 200 when the company reports the increase on Tuesday. The new users span 15 countries and include organizations in power, healthcare and telecommunications; Anthropic said the cohort includes firms and non-profits that produce critical programming code but did not identify them. Anthropic has said public release of Mythos carries significant risk; separate reports say the EU cybersecurity agency will receive access.