On SWE-Bench Verified, the model achieved a score of 70.6%. This performance is notably competitive when placed alongside significantly larger models; it outpaces DeepSeek-V3.2, which scores 70.2%, ...

The mission of Firehouse is to educate and inspire firefighters so that they can protect their communities and keep themselves safe. By providing news, lessons learned, tactical changes and focusing ...

New 2026 Delta Media AI Survey shows AI is ubiquitous in real estate, with 97% of brokers saying agents use AI as the ...

The International Fact-Checking Network (IFCN) leverages the expertise of its team, Poynter faculty, signatories and global partners to provide fact-checkers with the tools and resources they need to ...

第一，智能体部署的性价比超高：仅激活 30 亿参数，即可实现媲美激活参数量高出 10–20 倍模型的性能，为智能体部署提供极高的性价比。（达到了Sonnet4.5的水平。） 其次，长程推理、工具调用能力出色。通过精心设计的训练方案，该模型在长程推理、复杂工具调用以及执行失败后的恢复方面表现出色，确保在动态编码任务中具备稳健性能。 第三，集成方式也很灵活。适配多种 CLI ...