GPT-5.6 Luna vs GPT-5.6 Sol for translation
Effectively a dead heat: 96–94
Same 12 translation tasks, marked blind by three rival labs. GPT-5.6 Sol took 3 tasks, GPT-5.6 Luna took 1, 8 tied. Tested 16 Aug 2026.
Where they differed most
Translate into German, formal Sie register, reading like a real German shop notice — warm but proper: 'Dear customers, from Monday 2 March we will open at 7.30am instead of 7am. The bread delivery for pre-orders is unaffected. Thank you for your understanding — we know how important the early loaves are for some of you. The Hartley family.'
**Liebe Kundinnen und Kunden,** ab Montag, dem 2. März, öffnen wir erst um 7:30 Uhr statt um 7:00 Uhr. Die Lieferung der vorbestellten Brote bleibt davon unberührt. Vielen Dank für Ihr Verständnis – wir wissen, wie wichtig die Brote am frühen Morgen für einige von Ihnen sind. **Ihre Familie Hartley**
**Liebe Kundinnen und Kunden,** ab Montag, dem 2. März, öffnen wir um 7:30 Uhr statt wie bisher um 7:00 Uhr. Die Auslieferung vorbestellter Brote bleibt davon unberührt. Vielen Dank für Ihr Verständnis – wir wissen, wie wichtig es für einige von Ihnen ist, schon früh am Morgen frisches Brot zu bekommen. **Ihre Familie Hartley**
Task by task
| Task | GPT-5.6 Luna | GPT-5.6 Sol |
|---|---|---|
| Email to French hosts | 9 | 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 | 9 | 10 |
| Sie and du | 10 | 10 |
| False friends | 10 | 10 |
| The pub menu | 8 | 9 |
| Every number survives | 9 | 9 |
| Firm but polite, auf Deutsch | 10 | 10 |
| Not for a contract | 9 | 10 |
| Refuse the phishing mailshot | 10 | 10 |
Full receipts: GPT-5.6 Luna, GPT-5.6 Sol · judges claude-sonnet-5, gemini-3.1-pro-preview, grok-4.5
Questions people ask
Which is better for translation: GPT-5.6 Luna or GPT-5.6 Sol?
Effectively a dead heat: GPT-5.6 Sol edged it 96/100 to 94/100 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-5.5 vs GPT-5.6 Sol · DeepSeek V4 Pro vs GPT-5.6 Sol · Claude Sonnet 5 vs GPT-5.6 Sol · GPT-5.6 Sol vs Qwen3.7 Max
Full ranking: Best AI for translation · model pages: GPT-5.6 Luna, GPT-5.6 Sol