GitHub 实测了 Copilot 智能体框架，五个基准上不输原生，还省 token，支持 20 多种模型，值得试试。
GitHub 在 SWE-bench Verified、SWE-bench Pro、SkillsBench、TerminalBench、Win-Hill 五个基准上对 Copilot agentic harness 进行了对比测试。固定模型和任务后，任务解决率与模型原生 harness 持平。在大多数配置下 token 消耗更少，最高可省 30%。Copilot 支持超过 20 个模型，用户可针对任务自由切换效率或质量。
We benchmarked the GitHub Copilot agentic harness against the harnesses that ship leading models nat...
We benchmarked the GitHub Copilot agentic harness against the harnesses that ship leading models natively. Holding the model and task fixed across SWE-bench Verified, SWE-bench Pro, SkillsBench, TerminalBench, and Win-Hill, the results were clear: ✅ Task resolution on par with model-vendor harnesses ✅ Fewer tokens across most configurations 💡 A key learning: With GitHub Copilot supporting more than 20 models, you're free to pick efficiency or peak quality per task. 💬 2 🔄 1 ❤️ 19 👀 5894 📊 4 ⚡