GPT-5.3-Codex vs Kimi K3 for bookkeeping
GPT-5.3-Codex wins, 92–84
Same 12 bookkeeping tasks, marked blind by three rival labs. GPT-5.3-Codex took 9 tasks, Kimi K3 took 1, 2 tied. Tested 27 Aug 2026.
Where they differed most
A bank reconciliation will not balance. The bank statement closing balance is £8,412.55. The cash book closing balance is £8,376.55. Outstanding items already accounted for: an unpresented cheque of £120.00 and a lodgement of £120.00 not yet on the statement, which cancel each other out. The bookkeeper has checked her additions twice. The cash book receipts for the month, as entered: £451.00, £1,208.30, £96.75, £2,340.00, £415.00, £188.50. The corresponding bank credits: £415.00, £1,208.30, £96.75, £2,340.00, £451.00, £188.50. Work out the difference, say what kind of error it is, explain the arithmetic rule that identifies this class of error at a glance, and say which side is wrong and how she would prove it. Then say what she should do about it in the ledger and what she should NOT do.
You have an **unexplained difference of £36.00**:
\[
£8,412.55 - £8,376.55 = £36.00
\]
The £120 unpresented cheque and £120 lodgement cancel out, so they are **not** the cause.
### What kind of error this is
This is the classic pattern of a **transposition error** (digits swapped), e.g. **£451** vs **£415**.
\[
451 - 415 = 36
\]
### The quick arithmetic rule
If a reconciliation difference is divisible by **9**, suspect a transposition or similar digit error.
\[
36 \div 9 = 4 \text{ exactly}
\]
So £36 immediately points to that class of e…### 1. Difference
\[
£8,412.55-£8,376.55=\boxed{£36.00}
\]
The unpresented cheque and outstanding lodgement cancel:
\[
£8,412.55-£120+£120=£8,412.55
\]
So the unexplained difference remains **£36, with the bank balance higher than the cash book**.
### 2. Type of error indicated
£36 has the classic signature of a **digit-transposition error**:
\[
£451-£415=£36
\]
The rule is that a difference caused by transposing digits is divisible by 9:
\[
36\div 9=4
\]
For adjacent digits \(a\) and \(b\):
\[
(10a+b)-(10b+a)=9(a-b)
\]
More general…Task by task
| Task | GPT-5.3-Codex | Kimi K3 |
|---|---|---|
| The client lunch and the sandwich | 9 | 8 |
| Categorise, and admit what you cannot | 9 | 8 |
| The invoice totals | 10 | 10 |
| Find the £36 | 8 | 10 |
| Profit but no money | 10 | 8 |
| Not your decision to make | 10 | 9 |
| The faded shoebox | 8 | 8 |
| It is my company, it is my money | 9 | 8 |
| The mileage claim | 10 | 9 |
| Capital or revenue | 8 | 6 |
| Quote, invoice, receipt | 9 | 8 |
| Refuse the backdated invoice | 10 | 9 |
Full receipts: GPT-5.3-Codex, Kimi K3 · judges claude-sonnet-5, gemini-3.1-pro-preview, grok-4.5
Questions people ask
Which is better for bookkeeping: GPT-5.3-Codex or Kimi K3?
GPT-5.3-Codex — it scored 92/100 against 84/100 on our 12-task bookkeeping suite, winning 9 tasks to 1 with 2 tied. Every answer was marked blind by three judges from three rival AI labs.
How was this tested?
Both models answered the identical published bookkeeping tasks. Three AI judges from three different labs scored every answer blind against a fixed rubric; mechanically checkable rules (word limits, banned phrases) are enforced by the test harness in code. The raw outputs and judge verdicts are downloadable.
More bookkeeping head-to-heads: GPT-5.3-Codex vs GPT-5.6 Terra · GPT-5.6 Terra vs Kimi K3 · GPT-5.3-Codex vs GPT-5.6 Sol · GPT-5.6 Sol vs Kimi K3 · GPT-5.3-Codex vs GPT-5.5 · Claude Fable 5 vs GPT-5.3-Codex
Full ranking: Best AI for bookkeeping · model pages: GPT-5.3-Codex, Kimi K3