Claude Sonnet 5 vs GPT-5.6 Sol for translation

The verdict

GPT-5.6 Sol wins, 96–91

Same 12 translation tasks, marked blind by three rival labs. GPT-5.6 Sol took 4 tasks, Claude Sonnet 5 took 2, 6 tied. Tested 16 Aug 2026.

Where they differed most

The task — Idioms, not word-for-word

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.

Claude Sonnet 55/10

Voici les équivalents naturels :

1) **Il pleut des cordes.**
(It's raining ropes.)

2) **Cette voiture a coûté les yeux de la tête.**
(That car cost the eyes from the head.)

3) **Je ne suis pas dans mon assiette.**
(I'm not in my plate.)

4) **Il a vendu la mèche.**
(He sold the wick.)

GPT-5.6 Sol10/10

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 from my head.]  
3) Je ne suis pas dans mon assiette. [I’m not in my plate.]  
4) Il a vendu la mèche. [He sold the wick.]

Task by task

Task	Claude Sonnet 5	GPT-5.6 Sol
Email to French hosts	8	9
The B&B welcome note	9	9
Shop notice, formal German	10	9
French into real English	10	10
Idioms, not word-for-word	5	10
Sie and du	10	10
False friends	10	10
The pub menu	8	9
Every number survives	10	9
Firm but polite, auf Deutsch	9	10
Not for a contract	10	10
Refuse the phishing mailshot	10	10

Full receipts: Claude Sonnet 5, GPT-5.6 Sol · judges gemini-3.1-pro-preview, gpt-5.6-terra, grok-4.5

Questions people ask

Which is better for translation: Claude Sonnet 5 or GPT-5.6 Sol?

GPT-5.6 Sol — it scored 96/100 against 91/100 on our 12-task translation suite, winning 4 tasks to 2 with 6 tied. Every answer was marked blind by three judges from three rival AI labs.

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-5.6 Sol vs GPT-5.6 Terra · GPT-5.3-Codex vs GPT-5.6 Sol · GPT-5.5 vs GPT-5.6 Sol · DeepSeek V4 Pro vs GPT-5.6 Sol · GPT-5.6 Sol vs Qwen3.7 Max

Full ranking: Best AI for translation · model pages: Claude Sonnet 5, GPT-5.6 Sol