Will the tables inside the <think> block end up in my workbook?
No. The reasoning block is discarded before tables are detected, so only the tables in the answer are listed and written.
Paste a DeepSeek table — or the whole answer, reasoning block and all — and copy it into Excel as real cells, or download a .xlsx with the numbers typed. The <think> block is ignored, so its draft tables never end up in the workbook.
Select the table in DeepSeek, copy, and paste it into the box above. If the copy included the <think> reasoning, it is ignored, and only the tables in the answer are detected. Click Copy for Sheets / Excel and paste into Excel for separate columns with numbers recognised, or Download .xlsx for a typed workbook with a bold frozen header and one sheet per table. Nothing is uploaded.
Use the copy button under the answer or select the text. DeepSeek's copy gives markdown text, sometimes with the reasoning block in front; either form is read correctly here.
The table is detected as you paste and shown as a grid. Switch tables with the tabs if the answer had more than one.
Click Copy for Sheets / Excel and paste into Excel, or Download .xlsx for a typed file.
A DeepSeek thinking model drafts before it answers, and the drafts are often tables: a first attempt at the comparison, a reworked version with a column added, then the final table in the answer. When the copy includes the <think> block — a drag-select from above the answer, the API, a local model, a third-party client — all of those drafts come along. A table detector that reads the whole paste finds three tables where you wanted one, with the wrong one first.
This page discards a <think>…</think> block before it looks for tables, so only the tables in the answer itself are detected and listed in the tabs above the preview. Paste the whole answer without trimming; the reasoning, the prose and the markdown around the tables are ignored, and what is left is the grid DeepSeek actually gave you.
DeepSeek's tables are well-formed markdown: outer pipes, a dashes row under the header, a bold lead-in in the first column, and values that tend to be plain numbers rather than prose. That makes them good candidates for typing. Thousands separators, a leading $, € or £, a trailing % and negatives in parentheses or with a minus are read as numbers; the **bold** is removed from the cells; the header row is written bold and frozen in the .xlsx. Figures with units attached — 12 kg, 3–5 days — stay text, deliberately, and "numbers only, units in the header" in the prompt fixes that at the source.
Two DeepSeek specifics to watch. A quantitative table sometimes carries LaTeX in a cell — $\frac{1}{2}$ or \times 10^3 — which stays as text; ask for plain-text maths if the column needs to compute. And a full-width Chinese comma or colon occasionally appears where an ASCII one was meant, which the number detection does not read as a thousands separator; the preview shows it, and the cell is quick to fix.
DeepSeek's longer answers to analytical questions hold several tables with headings and prose between them. Paste the whole answer. Each table is detected separately; the tabs switch between them; Copy for Sheets / Excel and the CSV download act on the one selected; Download .xlsx writes every table as its own sheet, in order. For a workbook you intend to build on, the .xlsx route is usually fastest — numbers typed, headers bold and frozen, columns sized to their content — and it is built entirely in your browser.
DeepSeek follows formatting instructions well. "Give me the final table as CSV, header row first, numbers only, units in the column names, no LaTeX" produces data that types perfectly and pastes as columns even without this page. Add "no reasoning" if your client supports a non-thinking mode, and the <think> block never appears. Either way, the paste here reads what it is given — CSV, markdown pipes or an ASCII grid — and hands Excel a table.
No. The reasoning block is discarded before tables are detected, so only the tables in the answer are listed and written.
Yes, where they are unambiguous: thousands separators, currency symbols, percentages and negatives are typed in the .xlsx and right-aligned in the preview. Figures with units or LaTeX stay text.
DeepSeek occasionally emits full-width punctuation from the Chinese layout. It is not read as a thousands separator; fix the cell in the preview or in Excel and the value types.
Yes. Comma- and semicolon-separated text is parsed by the CSV rules, quoted fields included, and then copied or downloaded like any other table.
No. The conversion runs in your browser with client-side JavaScript. The DeepSeek answer is never sent to a server, stored or logged.