GPT-5.3-Codex vs GPT-5.6 Terra for translation
Effectively a dead heat: 94–93
Same 12 translation tasks, marked blind by three rival labs. GPT-5.6 Terra took 3 tasks, GPT-5.3-Codex took 1, 8 tied. Tested 16 Aug 2026.
Where they differed most
Give the natural French equivalent a French speaker would actually say for each — not literal translations: 1) 'It's raining cats and dogs.' 2) 'That car cost an arm and a leg.' 3) 'I'm feeling a bit under the weather.' 4) 'He let the cat out of the bag.' For each, one line: the French, then the literal back-translation in brackets so the difference is visible.
1) **Il pleut des cordes.** [“It’s raining ropes.”] 2) **Cette voiture m’a coûté les yeux de la tête.** [“That car cost me the eyes out of my head.”] 3) **Je suis un peu patraque.** [“I’m feeling a bit off/rough.”] 4) **Il a vendu la mèche.** [“He sold the wick.”]
1) Il pleut des cordes. [It’s raining ropes.] 2) Cette voiture a coûté les yeux de la tête. [That car cost the eyes out of one’s head.] 3) Je ne me sens pas très bien. [I don’t feel very well.] 4) Il a vendu la mèche. [He sold the wick.]
Task by task
| Task | GPT-5.3-Codex | GPT-5.6 Terra |
|---|---|---|
| Email to French hosts | 8 | 9 |
| The B&B welcome note | 9 | 9 |
| Shop notice, formal German | 10 | 10 |
| French into real English | 9 | 9 |
| Idioms, not word-for-word | 10 | 8 |
| Sie and du | 10 | 10 |
| False friends | 10 | 10 |
| The pub menu | 9 | 9 |
| Every number survives | 8 | 9 |
| Firm but polite, auf Deutsch | 10 | 10 |
| Not for a contract | 9 | 10 |
| Refuse the phishing mailshot | 10 | 10 |
Full receipts: GPT-5.3-Codex, GPT-5.6 Terra · judges claude-sonnet-5, gemini-3.1-pro-preview, grok-4.5
Questions people ask
Which is better for translation: GPT-5.3-Codex or GPT-5.6 Terra?
Effectively a dead heat: GPT-5.6 Terra edged it 94/100 to 93/100 on our translation suite — too close to matter, so pick on price or the product you already use.
How was this tested?
Both models answered the identical published translation 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 translation head-to-heads: GPT-5.6 Sol vs GPT-5.6 Terra · GPT-5.3-Codex vs GPT-5.6 Sol · GPT-5.6 Luna vs GPT-5.6 Terra · GPT-5.3-Codex vs GPT-5.6 Luna · GPT-5.5 vs GPT-5.6 Terra · DeepSeek V4 Pro vs GPT-5.6 Terra
Full ranking: Best AI for translation · model pages: GPT-5.3-Codex, GPT-5.6 Terra