Gemini 4 Argon for small business accounting and bookkeeping: one setting took it to 40 of 40

Gemini 4 Argon can review a month of small-business books, with one setting turned on. On our 40-check test month it scored 40 of 40 with thinking set to high, for 27 cents. With no thinking setting it scored 35 and 33 and handed back 5 to 7 checks as questions the data already pointed to. It never guessed a payment it couldn't place.

Owners already click through enough. On Xero's Product Ideas board, under a request for bank rules to apply automatically, Robert Passlow wrote: "Huge clicking OK 900 times, is not how I wanted to spend Friday."¹ QuickBooks Online owners know the same pile as the For Review tab. An AI that hands lines back as questions adds to it, and you end up babysitting the AI that was meant to clear it.

We build Booke AI, AI bookkeeping software for QuickBooks Online and Xero. The setting and the five prompts we used are below, ready to copy.

What is Gemini 4 Argon?

Gemini 4 Argon is Google's latest frontier model, released to developers and businesses on October 13, 2026. Google says it "excels at deep work in complex, real-world tasks across domains including coding, finance," and that it "answers questions more directly with less hedging."²

What matters for an owner:

• Where it runs: for businesses, Google lists Gemini for business, the paid tier of the Gemini API in AI Studio, the Gemini platform and Google Antigravity; Google AI Ultra and Pro subscribers get it in the Gemini app and AI Mode in Search.²

• Price: an introductory $2 per million input tokens and $10 per million output tokens; after November 30, 2026, $4 and $20.² Tokens are the small chunks of text a model reads and writes, and you pay for both.

• Context: it reads up to about 1 million tokens at once,³ far more than a month of bank lines.

Sundar Pichai's launch post passed 3.9 million views in under a day,⁴ and Demis Hassabis showed Argon off with a wizard chess game.⁵ We tested the finance part on the month a small-business owner closes.

Thinking on or off: the setting that decides the score

Thinking is the model working through the problem before it answers. We ran Argon through OpenRouter, a service that resells access to many AI models, where the setting is called reasoning effort.

• No thinking setting sent (two runs): zero reasoning tokens, 35 and 33 of 40, 5 and 7 extra questions, $0.0587 and $0.0618 for the month.

• Reasoning effort set to high (one run): 20,305 reasoning tokens, 40 of 40, no extra questions on our checks, $0.2738 for the month.

The 21 cents bought back every check the no-thinking runs handed back. On our month, the "less hedging" Google promises showed up only after we paid for thinking.

How Gemini 4 Argon scored against four other models

The month is a made-up September for a small candle business on QuickBooks Online: five bookkeeping jobs and 40 checks we wrote before any model saw the data. The same five prompts and 40 checks have now run on five models:

A wrong call is a check answered with a wrong account or amount, a guess where the data can't tell, a planted problem missed, or a false alarm. An extra question is a check the model handed back as a question when the data already pointed to the answer, like what a candle shop buys from a candle-supply company. Argon made no wrong calls in 120 checks across three runs; DeepSeek V4.1 Flash made one, in one of its runs.⁶

The setting and the prompts, ready to copy

Turn thinking on before you paste anything:

Thinking on. On OpenRouter, set reasoning effort to high. Other tools name it differently: find the thinking or reasoning setting and set it to high.

Every job uses the same rules block, pasted after the job prompt and before your data:

Review only. Use only the data I paste. Don't post, change or delete anything in my books. Don't guess what a purchase was for from the merchant name alone: if you can't tell, mark it ASK and write one short question I can answer. Flag anything that needs my accountant instead of deciding tax treatment.

Paste a context block with your data: your business and name, every account ending (business, personal and loans), your chart of accounts, which is the list of accounts your books sort money into, and short notes on your loan payment and the invoices customers still owe you. Ours:

Business: Fernwood Candle Co. LLC, single-member LLC, owner Maya Rivera. Books in QuickBooks Online. Month: September 2026.

Accounts: Business Checking ...2210; Business Savings ...4411; Chase Ink business card ...9031; owner's personal checking ...7790; owner's personal Visa ...5512; SBA loan #0034.

Chart of accounts: Sales; Wholesale Sales; Refunds; Materials; Shipping & Packaging; Advertising; Software & Subscriptions; Merchant Fees; Office Supplies; Travel; Meals; Interest Expense; Owner's Draw; Owner Contribution; SBA Loan; Chase Ink Card; Business Savings; Accounts Receivable.

SBA loan #0034 statement, payment 2026-09-05: total 1,150.00 = principal 980.40 + interest 169.60.

Open invoices in QuickBooks: INV-1043, Okafor Gift Shop (contact: J. Okafor), 750.00, due 2026-09-30.

Add one line about owner money, from our owner contribution vs loan test, so the call on money between you and the business stays yours:

Owner money: money I move into the business from my personal accounts is an owner contribution unless I say it's a loan. Zelle payments to me are owner's draws.

The five bookkeeping jobs and what Gemini 4 Argon returned

1. Sort owner money, loan payments and transfers

For each checking line, say what it is: income, expense, owner contribution, owner's draw, loan payment (split principal and interest if the data shows it), card payment, transfer between my accounts, or payment of an existing invoice. Use my notes. Mark ASK when the data doesn't say.

Six checking lines moved $13,812.18 in September, and only $169.60 of it is an expense, the loan interest. No run put any of the six in the wrong place: the $5,000 from the owner's personal checking to Owner Contribution, the $1,150 SBA payment split $980.40 to the loan and $169.60 to interest, the card autopay as a card payment and the savings move as a transfer, neither one an expense, the $2,000 Zelle to the owner as Owner's Draw, and the $750 Zelle from J Okafor applied to the open invoice INV-1043 "rather than recording as new sales income so revenue isn't double-counted" (run 1, no thinking setting). Without thinking, run 1 hedged on the Zelle to the owner and run 2 also asked whether the $5,000 was a loan. The owner-money line above settles both: in our October 9 test, 5 of 6 runs on three other models asked nothing about owner money once it was in the notes.

2. Clear the Uncategorized pile

I keep my books in QuickBooks Online. Below are bank lines sitting in Uncategorized, my business context and my chart of accounts. For each line give: date, description, amount, suggested account from my chart (or TRANSFER with the account it moves to, or ASK, or ACCOUNTANT), and a one-line reason. Lines that move money between my own accounts or pay a loan or card are not expenses or income. End with the list of questions for me.

On Xero, name Xero and paste your unreconciled bank lines instead. Every run caught the $1,500 savings move and the $3,412.18 card autopay as transfers; booked as expenses, they would have overstated September's costs by $4,912.18. The $2,200 IRS payment went to ACCOUNTANT. Without thinking, both runs asked about three vendor lines, starting with the $642.18 CandleScience order: "CandleScience sells both candle-making supplies and packaging, so I cannot tell from the vendor name alone whether this belongs in Materials or Shipping & Packaging" (run 1). With thinking on, CandleScience went to Materials, Shopify to Software & Subscriptions and USPS to Shipping & Packaging.

The $400 Venmo payment, the $89.47 Amazon order and the $1,850 mobile deposit came back as questions in every run. A model that books that deposit as Sales invents $1,850 of income that may be your own money.

3. Explain why Stripe paid you less than you sold

My Stripe sales are higher than what reached my bank. Using the Stripe balance export and my bank deposits, explain the difference line by line: gross sales, card fees, refunds, disputes and any other deductions, per payout. Check that each payout equals a bank deposit. Show the totals I should record so that sales, fees and refunds are separate. Tell me anything I need to act on.

The $213.50 gap between $1,072.00 sold and $858.50 deposited closed in every run: $33.50 in card fees, a $45.00 refund, a $120.00 dispute (a customer's chargeback) and a $15.00 dispute fee. Both payouts matched their deposits, and every run gave $1,072.00 of gross sales as the number to record. Run 1 without thinking told the owner to "Check your Stripe dashboard immediately to submit evidence if you wish to contest the dispute".

4. Run the month-end checks

Check my September activity against earlier months and my receipt list. Find: possible duplicate charges, subscriptions whose price changed, and charges of 75 dollars or more with no receipt. Don't flag repeat purchases as duplicates unless the evidence points to one order charged twice. Give a short to-do list.

Every run caught CandleScience order 48650 charged twice for $312.40, Google Workspace rising from $14.00 to $16.80, Shopify from $39.00 to $105.00, and $228.75 of Staples and WebstaurantStore spending with no receipts. None flagged the two traps we planted: two $12.40 USPS labels with a receipt each, and an $18.20 charge under the $75 line.

5. Split a personal card you used for the business

I paid some business costs with my personal card. For each line, mark BUSINESS, PERSONAL or ASK, with the account from my chart for business lines. Treat refunds as reducing the cost they refund. Then give the net total of business lines that I should record as paid with personal funds, and the questions for me.

All three runs landed on $219.95 of business costs paid personally, with a $35.00 CandleScience refund netted against its order, and asked about Whole Foods, Shell, Delta and Target. Without thinking, run 1 also asked about Spotify and run 2 about Netflix and Spotify; with thinking on, both came back as personal.

How to use Gemini 4 Argon on your own month

1. Pick a month you've already closed, so you can check every answer against what you booked.

2. Export that month's bank and card lines from QuickBooks Online or Xero with date, description, amount and account.

3. Write the context block and notes: your business, every account ending, your chart of accounts, loan splits, open invoices and the owner-money line.

4. Turn thinking on, then paste one job at a time: the job prompt, the rules block, then the data. Start with owner money.

5. Check every line the model moved and every total, answer its questions, and make the changes in QuickBooks or Xero yourself.

How we tested

The month is synthetic: September 2026 for Fernwood Candle Co. LLC, a one-owner candle business on QuickBooks Online. We wrote the 40 checks on October 4, 2026 and no model ever saw them. Every model got the same prompts and data, pasted as text, with no connection to QuickBooks or a bank. Gemini 4 Argon ran three times on October 14 through OpenRouter, served by Google: twice with no thinking setting sent and once with reasoning effort set to high. Costs are at Google's introductory price. The thinking run slipped once outside our checks, writing that a $45.00 refund "reduced your net payout by $46.61 overall"; the $1.61 fee had already come off the sale. The full write-up is on X.⁷

What 40 of 40 still leaves you to do

The five steps above are the babysitting Argon still needs every month, even at high. The way we ran it, it sees only what you paste and writes nothing to your books: its perfect run ended with 40 right answers on a screen and none of them in QuickBooks. The model was the cheap part, 27 cents. The rest is export, paste, check and type it back.

Where Booke fits

Booke AI takes the export and the paste out of that routine. It works every day inside the QuickBooks Online or Xero books you already have, on the bank feeds you've connected, with no new platform and no new bank connection. It categorizes and matches the feed using your books' history and automates up to 80% of the manual bank-feed work.

It brings you only the transactions that need you. Unclear ones, the kind Argon asked about like the Venmo payment and the $1,850 deposit, are flagged for your decision instead of being booked as done, and the changes you approve improve future automation. Booke AI costs $129 per business per month.

FAQ

Can Gemini 4 Argon do my bookkeeping?

It can do the monthly review. With thinking on, it matched all 40 checks on our test month across owner money, the Uncategorized pile, Stripe payouts, month-end checks and a personal card. Run on pasted data, as we ran it, it posts nothing to QuickBooks Online or Xero, so you make every change yourself.

How does Gemini 4 Argon compare with GPT-6.1 Sol and Claude?

On the same 40 checks, Argon with thinking on matched 40 with no extra questions. GPT-6.1 Sol matched 30 and asked 10 extra questions; Claude Haiku 5.5 matched 33 to 34 and asked 6 to 7. Meta Muse Wonder is the only other model to reach 40, on its second run. We haven't run Claude Opus 5.5 or Sonnet 5.5 on this month.

How much does Gemini 4 Argon cost for a month of bookkeeping?

Our test month cost 27 cents with thinking on and 6 cents without, at Google's introductory price of $2 per million input tokens and $10 per million output tokens.² Through OpenRouter, thinking is billed as output tokens. After November 30 the price doubles to $4 and $20, and the same month comes to about 55 cents with thinking and about 12 cents without.

Is Gemini 4 Argon free?

Not on the options Google lists for a business: the paid tier of the Gemini API, Gemini for business, and the Google AI Ultra and Pro subscriptions. Google's one free offer is for scientists: 10,000 of them can apply for six months of Antigravity's highest usage tier.²

Is Gemini 4 Argon out?

Yes. Google released it globally on October 13, 2026: for businesses in Gemini for business, the API and Google Antigravity, and for Google AI Ultra and Pro subscribers in the Gemini app and AI Mode in Search.² OpenRouter listed it the same day.³

Keep reading

• Meta Muse Wonder for small business bookkeeping: 40 of 40 on our test month

• Claude Haiku 5.5 for small business bookkeeping: we tested it on a month of books

• ChatGPT Finances for small business: what it sees, and 5 bookkeeping prompts we tested

• Owner contribution, loan or owner's draw: what AI did with an owner's $5,000

Sources

1. Xero Product Ideas, Bank reconciliation - Apply bank rules automatically by default, comment by Robert Passlow, June 13, 2025. Read October 14, 2026.

2. Google, Gemini 4 Argon: Built to solve your hardest problems, Koray Kavukcuoglu, October 13, 2026.

3. OpenRouter, Google: Gemini 4 Argon: providers, context length and listing date. Read October 14, 2026.

4. Sundar Pichai (@sundarpichai), Gemini 4 Argon is here, X, October 13, 2026.

5. Demis Hassabis (@demishassabis), wizard chess game built with Gemini 4 Argon, X, October 13, 2026.

6. Vadim Chumak (@vadimvchumak), DeepSeek V4.1 Flash on the same 40 checks, X, October 5, 2026.

7. Vadim Chumak (@vadimvchumak), Stop babysitting AI on your books. One setting got Gemini 4 Argon to 40 of 40, X, October 14, 2026.