Unverified80% confidenceFactTime unknown
在旅行规划任务中，2024年初的模型成功率接近0%，但在工具辅助下GPT-4和Claude 3可以达到90%以上的正确率
1
Sources
80%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
李宏毅2026课程笔记：AI Agent三大核心能力深度解析
bilibili李宏毅agent
Related Entities
Related Claims
UnverifiedThe per-step accuracy of current top-tier models is approximately between 90% and 95%.70% similarUnverifiedUsing a GPT-3.5-level model for simple tasks costs approximately 1/20 of flagship model pricing, and proper LLM routing can reduce overall inference costs by 60%–80%.67% similarUnverifiedA 2023 Stanford University study showed that multimodal models such as GPT-4V, when parsing flowcharts with nested conditional logic, had accuracy approximately 15–25 percentage points lower than equivalent textual descriptions.67% similarUnverified2023年人类在GAIA测试中的平均成功率约为90%，而当时最强的GPT-4在Level 1上仅勉强达到15%65% similarUnverified在Spider基准上，早期单模型方案准确率约为70%，GPT-4等大模型将其提升到80%以上，多智能体协作策略的系统已能达到85%甚至更高的执行准确率63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/26238API
curl https://kongchang.com/api/v1/knowledge/claims/26238MCP
get_claim(id=26238)