TranslateGemma 27B
TranslateGemma 27B is a translation-specialized foundation model developed by Google DeepMind. It has a parameter scale of approximately 29B and supports a long context window of up to 128K.
Parameters
29.0B
Context Window
128K
License
Gemma License
Release Date
2026-01-15
API Pricing
API pricing for this model is not yet available
Strengths
- ・High performance specialized in translation
- ・128K broad context window
- ・Developed by Google DeepMind
Weaknesses
- ・Huge 55GB file size
- ・Unknown adaptability to general tasks
- ・Subject to specific license restrictions
Use Cases
- ・High-quality multilingual translation
- ・Translating long documents
- ・Foundation for translation-specific AI
Deep Analysis
Parameters
27B
Largest and most capable TranslateGemma variant
Performance
Highest quality in family
Best translation quality among all TranslateGemma models
Languages
55 core / ~500 extended
Most comprehensive language coverage
Release Date
January 15, 2026
Part of TranslateGemma family launch
Multimodal
Yes
Best image text translation in the family
Deployment
Server / GPU required
Requires significant compute for deployment
Training
SFT + RL
Distilled from Gemini with reinforcement learning optimization
Strengths
- ・Highest translation quality in the TranslateGemma family
- ・Best multimodal (image text) translation performance
- ・55 core languages with broadest extended coverage
- ・Distilled from Gemini models for maximum fidelity
- ・Open weights for full customization
- ・Suitable for the most demanding translation tasks
Weaknesses
- ・Largest model, requires significant GPU resources
- ・Not suitable for mobile or edge deployment
- ・Marginal quality gain over 12B may not justify 2x compute cost
- ・Translation-focused, not general-purpose
- ・Extended language pairs lack confirmed evaluation metrics
Competitor Comparison
| Model | Arena | SWE | GPQA | Price |
|---|---|---|---|---|
| TranslateGemma 27B | N/A | N/A | N/A | Free (open weights) |
| TranslateGemma 12B | N/A | N/A | N/A | Free (open weights) |
| Gemma 3 27B (baseline) | ~1430 | N/A | ~70% | Free (open weights) |
| NLLB-200 (54.5B) | N/A | N/A | N/A | Free (open weights) |
| GPT-5 (general) | ~1480 | ~80% | ~90% | $5/$20 per 1M |
TranslateGemma 27B is the largest and most capable variant in Google's open translation model family. It delivers the highest translation quality across 55 languages with multimodal image text translation support. Best suited for demanding server-side translation workloads.
Sources
Analysis generated: 2026-05-24