# widesearch / widesearch-ws-zh-060 - taskset: [widesearch](https://harnessreport.com/tasks/widesearch.md) - difficulty: hard - category: information-retrieval - language: - runnable from the site: no - agent timeout: 1800s ## Results by harness _none yet_ ## Instruction ``` 想要看下最近的大模型表现,列举在gemini2.5 pro &lite & flash-lite(区分开thinking和不开thinking模式)、claude3.7及以后发布的模型(截止到2025年6月)、o3&o3mini&o4mini、doubao-1.5-thinking和1.6thinking、deepseek v3及以后发布的主要模型(以上所有模型截止到2025年6月) ,具体想了解的指标包括模型号(如gemini-2.5-pro)、上下文窗口(如32k)、aime-2025指标、swe verified指标(单次尝试)、tau-bench-retail和airline指标。请尽可能在对应模型官网上搜索所有指标,但不能编造,如过官网里没有发布,相关指标可以输出NA。请以一整个Markdown表格的格式输出整理后的数据,不要拆分成多个markdown表格,每个单元格都需要按列名要求输出,不得无故省略,输出采用中文。 表格中的列名依次为: 模型名称、公司、上下文窗口、AIME 2025、SWE-bench Verified、TAU-bench-Airline、Tau-bench-Retail。不要问我任何问题,只需输出结果,输出格式为```markdown{数据内容}``` --- **Important:** Write your final answer as a Markdown table to the file `/workspace/output.md`. The file should contain ONLY the Markdown table (with header row and separator row). Do not include any other text, explanation, or code blocks — just the raw Markdown table. ``` --- Harness Report runs agent harnesses from their GitHub repos on Harbor tasks and records every model call. Every page is also `.md` and `.json`; index: https://harnessreport.com/llms.txt · MCP: https://harnessreport.com/mcp