GPT-5.6 Sol vs GPT-5.6 Terra for bookkeeping

The verdict

Dead heat: 93 each

Same 12 bookkeeping tasks, marked blind by three rival labs. Tested 18 Aug 2026.

Where they differed most

The task — Categorise, and admit what you cannot

Nine lines from a sole trader's business bank feed. She is a self-employed dog groomer. 1) 'AMZN Mktp UK £41.99'. 2) 'PETSMART WHOLESALE £238.00'. 3) 'SHELL BRIGHOUSE £62.10'. 4) 'TRANSFER TO J CROSSLEY £600.00'. 5) 'SAGE UK LTD £22.00'. 6) 'B&Q 1147 £84.55'. 7) 'HMRC NDDS £310.00'. 8) 'COSTA COFFEE £9.40'. 9) 'DVLA VEHICLE TAX £190.00'. Use only these categories: Stock and consumables, Motor and travel, Equipment, Software and subscriptions, Premises, Drawings, Tax and NIC, Cannot tell. Reply with ONLY valid JSON — no prose before or after — an array of nine objects in the order given, each with exactly these keys and no others: line (number), description (string, copied exactly), category (string, one of the eight above), confident (boolean), question (string — the single question you would put to the client, or null if you have none).

GPT-5.6 Sol9/10

[
  {
    "line": 1,
    "description": "AMZN Mktp UK £41.99",
    "category": "Cannot tell",
    "confident": false,
    "question": "What was purchased from Amazon?"
  },
  {
    "line": 2,
    "description": "PETSMART WHOLESALE £238.00",
    "category": "Stock and consumables",
    "confident": true,
    "question": null
  },
  {
    "line": 3,
    "description": "SHELL BRIGHOUSE £62.10",
    "category": "Motor and travel",
    "confident": false,
    "question": "Was this fuel for a vehicle used for the dog-grooming business?"
  },
  {…

GPT-5.6 Terra10/10

[
  {
    "line": 1,
    "description": "AMZN Mktp UK £41.99",
    "category": "Cannot tell",
    "confident": false,
    "question": "What was purchased from Amazon and was it for the business?"
  },
  {
    "line": 2,
    "description": "PETSMART WHOLESALE £238.00",
    "category": "Stock and consumables",
    "confident": true,
    "question": null
  },
  {
    "line": 3,
    "description": "SHELL BRIGHOUSE £62.10",
    "category": "Motor and travel",
    "confident": false,
    "question": "Was this fuel purchased for business travel?"
  },…

Task by task

Task	GPT-5.6 Sol	GPT-5.6 Terra
The client lunch and the sandwich	8	8
Categorise, and admit what you cannot	9	10
The invoice totals	10	10
Find the £36	9	8
Profit but no money	8	9
Not your decision to make	10	9
The faded shoebox	10	10
It is my company, it is my money	8	8
The mileage claim	10	10
Capital or revenue	10	10
Quote, invoice, receipt	9	10
Refuse the backdated invoice	10	10

Full receipts: GPT-5.6 Sol, GPT-5.6 Terra · judges claude-sonnet-5, gemini-3.1-pro-preview, grok-4.5

Questions people ask

Which is better for bookkeeping: GPT-5.6 Sol or GPT-5.6 Terra?

They tied at 93/100 on our bookkeeping suite — for this job, pick on price or the product you already use.

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-5.5 vs GPT-5.6 Terra · Claude Fable 5 vs GPT-5.6 Terra · GPT-5.6 Terra vs Grok 4.5 · GPT-5.6 Terra vs Qwen3.7 Max · GPT-5.6 Terra vs Kimi K3

Full ranking: Best AI for bookkeeping · model pages: GPT-5.6 Sol, GPT-5.6 Terra