Datasets:
| {"id": "t-01", "tool": "browser", "question": "Open https://tokenfactory.nebius.com/models (it renders with JavaScript). How many public model endpoints does Nebius Token Factory list there?", "answer": "28", "answer_type": "number", "tolerance": 0.0, "source": "https://tokenfactory.nebius.com/endpoints", "checked": "2026-10-18"} | |
| {"id": "t-02", "tool": "browser", "question": "On the Nebius Token Factory model endpoints page (https://tokenfactory.nebius.com/models), how many total parameters does Reflection Beam 1.0 have according to its description?", "answer": "501.3B", "aliases": ["501.3 billion", "501.3"], "answer_type": "text", "source": "https://tokenfactory.nebius.com/endpoints", "checked": "2026-10-18"} | |
| {"id": "t-03", "tool": "browser", "question": "On the Nebius Token Factory model endpoints page (https://tokenfactory.nebius.com/models), which processing region is shown for DeepSeek-V4-Pro-0813?", "answer": "uk-south1", "answer_type": "text", "source": "https://tokenfactory.nebius.com/endpoints", "checked": "2026-10-18"} | |
| {"id": "t-04", "tool": "extract", "question": "Nebius publishes its Token Factory catalog as JSON at https://tokenfactory.nebius.com/api/public/models_info. Which NVIDIA Nemotron model in it has the highest tokens per second?", "answer": "Nemotron-3-Ultra-550b-a55b", "aliases": ["Nemotron 3 Ultra", "Nemotron-3-Ultra", "Nemotron Ultra"], "answer_type": "text", "source": "https://tokenfactory.nebius.com/api/public/models_info", "checked": "2026-10-18"} | |
| {"id": "t-05", "tool": "extract", "question": "In the Nebius Token Factory catalog JSON at https://tokenfactory.nebius.com/api/public/models_info, what is the output price per 1M tokens of nvidia/Nemotron-3_5-Lightning, in US dollars?", "answer": "0.24", "answer_type": "number", "tolerance": 0.0, "source": "https://tokenfactory.nebius.com/api/public/models_info", "checked": "2026-10-18"} | |
| {"id": "t-06", "tool": "map/crawl", "question": "Using the Tavily documentation at docs.tavily.com, how many API credits does one Tavily Search request with search_depth 'advanced' cost?", "answer": "2", "answer_type": "number", "tolerance": 0.0, "source": "https://docs.tavily.com/documentation/api-credits", "checked": "2026-10-18"} | |
| {"id": "t-07", "tool": "map/crawl", "question": "Using the Tavily documentation at docs.tavily.com: how many API credits in total do 3 advanced searches plus one basic Extract call that successfully extracts 10 URLs cost?", "answer": "8", "answer_type": "number", "tolerance": 0.0, "source": "https://docs.tavily.com/documentation/api-credits", "checked": "2026-10-18", "computed": "3 x 2 + 10/5 x 1"} | |
| {"id": "t-08", "tool": "map/crawl", "question": "According to docs.tavily.com, how many credits does a regular Tavily Map call cost per 10 successfully returned pages?", "answer": "1", "answer_type": "number", "tolerance": 0.0, "source": "https://docs.tavily.com/documentation/api-credits", "checked": "2026-10-18"} | |
| {"id": "t-09", "tool": "code", "question": "In the GitHub repository https://github.com/tavily-ai/tavily-python, what is the default value of the timeout parameter of TavilyClient.search, in seconds?", "answer": "60", "answer_type": "number", "tolerance": 0.0, "source": "https://github.com/tavily-ai/tavily-python/blob/master/tavily/tavily.py", "checked": "2026-10-18"} | |
| {"id": "t-10", "tool": "code", "question": "In the GitHub repository https://github.com/sami-mag07/searchify-l, which environment variable selects the reranker model?", "answer": "RERANKER_MODEL", "answer_type": "text", "source": "https://github.com/sami-mag07/searchify-l/blob/main/src/searchify/rerank.py", "checked": "2026-10-18"} | |
| {"id": "t-11", "tool": "code", "question": "In the GitHub repository https://github.com/ChromeDevTools/chrome-devtools-mcp, which command-line flag makes the server use a temporary user data directory that is cleaned up after the browser closes?", "answer": "--isolated", "aliases": ["isolated"], "answer_type": "text", "source": "https://github.com/ChromeDevTools/chrome-devtools-mcp", "checked": "2026-10-18"} | |
| {"id": "t-12", "tool": "search+extract", "question": "How many rows (questions) does the google/frames-benchmark dataset on Hugging Face have?", "answer": "824", "answer_type": "number", "tolerance": 0.0, "source": "https://huggingface.co/datasets/google/frames-benchmark", "checked": "2026-10-18"} | |
| {"id": "t-13", "tool": "search+browser", "question": "Reflection AI released an open-weight model in October 2026. What is its output price per 1M tokens on Nebius Token Factory, in US dollars?", "answer": "1.00", "aliases": ["$1", "1"], "answer_type": "number", "tolerance": 0.0, "source": "https://tokenfactory.nebius.com/endpoints", "checked": "2026-10-18"} | |
| {"id": "t-14", "tool": "any", "question": "What is the output price per 1M tokens of NVIDIA Nemotron 4 Mega on Nebius Token Factory?", "answer": "No such model is listed on Nebius Token Factory", "answer_type": "unanswerable", "source": "https://tokenfactory.nebius.com/api/public/models_info", "checked": "2026-10-18"} | |