GPT-5.3-Codex vs GPT-5.6 Luna for translation

The verdict

Effectively a dead heat: 9493

Same 12 translation tasks, marked blind by three rival labs. GPT-5.6 Luna took 3 tasks, GPT-5.3-Codex took 2, 7 tied. Tested 16 Aug 2026.

Where they differed most

The task — Email to French hosts

Translate into French for a couple in their 60s the writer has never met — vous throughout, natural French correspondence rather than word-for-word: 'Dear Monsieur and Madame Perrin, Thank you so much for having us in September. We will arrive in Limoges by train on Friday the 12th at around 6pm. My husband is allergic to shellfish, but otherwise we eat everything. Could we bring you anything from England that you miss? With best wishes, Janet.'

Monsieur et Madame Perrin,

Merci beaucoup de nous accueillir en septembre.  
Nous arriverons à Limoges en train le vendredi 12, vers 18 h.  
Mon mari est allergique aux fruits de mer, mais à part cela, nous mangeons de tout.  
Souhaitez-vous que nous vous apportions quelque chose d’Angleterre qui vous manque ?

Bien cordialement,  
Janet
Chers Monsieur et Madame Perrin,

Merci beaucoup de nous accueillir en septembre. Nous arriverons à Limoges en train le vendredi 12 vers 18 heures.

Mon mari est allergique aux fruits de mer, mais à part cela, nous mangeons de tout. Pourrions-nous vous apporter quelque chose d’Angleterre qui vous manque ?

Bien cordialement,  
Janet

Task by task

TaskGPT-5.3-CodexGPT-5.6 Luna
Email to French hosts89
The B&B welcome note99
Shop notice, formal German1010
French into real English910
Idioms, not word-for-word109
Sie and du1010
False friends1010
The pub menu98
Every number survives89
Firm but polite, auf Deutsch1010
Not for a contract99
Refuse the phishing mailshot1010

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 translation: GPT-5.3-Codex or GPT-5.6 Luna?

Effectively a dead heat: GPT-5.6 Luna 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 Luna vs GPT-5.6 Sol · GPT-5.3-Codex vs GPT-5.6 Sol · GPT-5.6 Luna vs GPT-5.6 Terra · GPT-5.5 vs GPT-5.6 Luna · DeepSeek V4 Pro vs GPT-5.6 Luna · Claude Sonnet 5 vs GPT-5.6 Luna

Full ranking: Best AI for translation · model pages: GPT-5.3-Codex, GPT-5.6 Luna