GPT-5.3-Codex vs GPT-5.6 Luna for humanising AI text
GPT-5.3-Codex wins, 97–92
Same 12 humanising AI text tasks, marked blind by three rival labs. GPT-5.3-Codex took 5 tasks, GPT-5.6 Luna took 0, 7 tied. Tested 16 Aug 2026.
Where they differed most
Rewrite for the school newsletter in a head teacher's human voice — warm, direct, every fact kept: the old date, the new date, the departure time, the wellies: 'Please be advised that due to unforeseen circumstances relating to ground conditions, the Year 4 farm visit has been rescheduled from Tuesday 8 October to Thursday 17 October. Departure will occur at 9.15am promptly. It is imperative that all pupils are equipped with wellington boots on the day in question. We apologise for any inconvenience caused and thank you for your continued understanding and support.'
Dear families, A quick update about the Year 4 farm visit: because of unforeseen ground conditions, we’ve had to move the trip from **Tuesday 8 October** to **Thursday 17 October**. Please make sure children are ready to leave **promptly at 9.15am** on the day, and that every pupil comes in **wellington boots**. We’re sorry for any inconvenience this change may cause, and thank you, as always, for your understanding and support.
**Year 4 Farm Visit – New Date** Due to unforeseen ground conditions, the Year 4 farm visit has been moved from **Tuesday 8 October** to **Thursday 17 October**. The coach will leave promptly at **9.15am**, so please ensure your child arrives on time and brings **wellington boots**. We apologise for the change and any inconvenience this may cause. Thank you, as always, for your understanding and support.
Task by task
| Task | GPT-5.3-Codex | GPT-5.6 Luna |
|---|---|---|
| The touching-base email | 10 | 9 |
| De-guru the bio | 10 | 9 |
| Product page, human voice | 10 | 9 |
| Strip the tics | 9 | 8 |
| The head's newsletter | 10 | 8 |
| The condolence card | 10 | 10 |
| Reply to the review, humanly | 10 | 10 |
| About us, actually about us | 10 | 10 |
| Best man, not best template | 9 | 9 |
| Human but still formal | 9 | 9 |
| Change log required | 9 | 9 |
| Refuse the disguise job | 10 | 10 |
Full receipts: GPT-5.3-Codex, GPT-5.6 Luna · judges claude-sonnet-5, gemini-3.1-pro-preview, grok-4.5
Questions people ask
Which is better for humanising AI text: GPT-5.3-Codex or GPT-5.6 Luna?
GPT-5.3-Codex — it scored 97/100 against 92/100 on our 12-task humanising AI text suite, winning 5 tasks to 0 with 7 tied. Every answer was marked blind by three judges from three rival AI labs.
How was this tested?
Both models answered the identical published humanising AI text 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 humanising AI text head-to-heads: Claude Sonnet 5 vs GPT-5.3-Codex · Claude Fable 5 vs GPT-5.3-Codex · GPT-5.3-Codex vs GPT-5.6 Sol · GPT-5.3-Codex vs GPT-5.5 · GLM 5.2 vs GPT-5.3-Codex · GPT-5.3-Codex vs GPT-5.6 Terra
Full ranking: Best AI for humanising AI text · model pages: GPT-5.3-Codex, GPT-5.6 Luna