Guide

How to Use DeepSeek in Excel

August 25, 2026 · 5 min read

DeepSeek is the cheapest serious AI you can put inside a spreadsheet, and unlike the flat-rate alternatives there is no waitlist: an API key works the minute you create it on the EBITDAI Pro plan, and a complete financial model build costs a few cents. DeepSeek ships no Excel add-in of its own, so you connect it through an add-in that accepts your own key. Here is the whole setup, the exact costs, and where DeepSeek falls short.

The five steps

1Install the EBITDAI add-inIn Excel, go to the Insert tab, choose Get Add-ins, and search for EBITDAI. It works in Excel on the web and on desktop, on both Windows and Mac.
2Create a DeepSeek account and top upSign in at platform.deepseek.com and add a small balance; DeepSeek requires a top-up before keys work, and a few dollars lasts most users months.
3Create an API keyUnder API keys, create a key named EBITDAI and copy it immediately; DeepSeek shows a key only once.
4Connect it in the task paneOpen the EBITDAI pane, choose DeepSeek on the Connect screen, and paste the key. One test call verifies it. The key is stored on your device, not on our servers.
5Ask for a model in plain EnglishType a request such as 'Build a 3-statement model for a SaaS business with $2M ARR growing 40%'. The agent reads your actual cells and writes the result into the workbook.

The full walkthrough with screenshots is in the API key tutorial.

What it costs, exactly

DeepSeek V4-Flash publishes rates of $0.22 per million input tokens and $0.66 per million output tokens off-peak, doubling to $0.44 and $1.32 at peak hours. Those numbers are abstract until you translate them: in our production benchmarks a full three-statement model build lands at a few cents, and a $10 usage pack covers 200 to 300 complete model builds. Most users never spend close to that in a month.

EBITDAI charges $15/month on Pro (with 7.5x more usage than Lite across every model) and adds zero markup on tokens; you pay DeepSeek directly at their published rates. There is no credit system and no meter other than DeepSeek's own balance page.

The privacy angle is real

Because DeepSeek's API is OpenAI-compatible, EBITDAI calls it directly from your machine. Your workbook content goes to DeepSeek and nowhere else; nothing routes through EBITDAI's servers, and your key never leaves your device. The details are on the security page.

Where DeepSeek falls short

Accuracy matters here. In our published benchmark of 10 real modeling tasks, EBITDAI on DeepSeek passed 26 of 30 runs at roughly $0.03 of model cost per task; Kimi K3 and Claude for Excel both passed 30 of 30. The misses concentrated in the hardest multi-tab tasks. If you build complex LBO models daily, a flat-rate Kimi Code subscription in Excel is the stronger pick; Kimi's waitlist is covered in the subscription guide. Many users run both: DeepSeek for everyday work, Kimi for the heavy builds, switched mid-conversation.

Frequently asked questions

Does DeepSeek have an official Excel add-in?

No. DeepSeek ships no Excel integration of its own. But its API is OpenAI-compatible, so an add-in that accepts your own key can connect to it. EBITDAI does exactly that: paste a DeepSeek key into the task pane and the agent reads and edits your workbook on your account.

How much does DeepSeek cost in Excel?

DeepSeek V4-Flash API rates are $0.22 per million input tokens and $0.66 per million output tokens off-peak, doubling to $0.44 and $1.32 at peak hours. In our production benchmarks a complete financial model build costs a few cents, so a $10 usage pack covers 200 to 300 full model builds. EBITDAI itself is $15/month on Pro, which includes 7.5x more usage than Lite across every model and adds zero markup to tokens billed on your own key. Own keys are a Pro feature.

Is there a waitlist?

No. A DeepSeek API key works the minute you create it. That makes DeepSeek the standard bridge while you wait for a Kimi Code subscription to clear its waitlist: start on DeepSeek today, add the Kimi key when approval lands, and EBITDAI switches models mid-conversation.

Is my spreadsheet data safe?

DeepSeek's API is called directly from your machine, so your workbook data goes to DeepSeek and nowhere else; nothing touches EBITDAI's infrastructure at all. Your API key is stored on your device, not on our servers.

How good is DeepSeek compared to Kimi k3 or Claude?

In our published benchmark of 10 real modeling tasks with 30 runs per tool, EBITDAI on DeepSeek passed 26 of 30 at roughly $0.03 of model cost per task, while Kimi K3 and Claude for Excel both passed 30 of 30. DeepSeek is the value pick: near-frontier quality at pay-per-cents pricing, with the misses concentrated in the hardest multi-tab tasks.

Cents per model, starting now

EBITDAI runs on the Pro plan at $15/month, with a free plan (150 requests a week on Meta Muse Spark 1.3, where Meta trains on everything you send) to start. No waitlist between you and your first build.

See how it works