Skip to content

Model gemma3 27b

hydropix edited this page Dec 23, 2025 · 1 revision

gemma3:27b

Ollama Model ID: gemma3:27b


Summary

Metric Value
Average Score 🟡 7.1/10
Accuracy 7.7/10
Fluency 7.1/10
Style 6.8/10
Languages Tested 19
Total Translations 95
Best Language Spanish (7.8)
Worst Language Bengali (5.8)

Language Performance

Top Languages

Rank Language Overall Accuracy Fluency Style
1 Spanish 🟡 7.8 8.6 7.8 7.8
2 Italian 🟡 7.8 8.0 7.8 7.2
3 Portuguese 🟡 7.6 8.2 7.6 7.2
4 Chinese (Traditional) 🟡 7.6 7.6 7.2 7.2
5 Russian 🟡 7.6 7.8 7.6 7.0
6 Arabic 🟡 7.6 8.0 7.6 7.4
7 French 🟡 7.4 7.4 7.6 7.0
8 Chinese (Simplified) 🟡 7.4 7.6 7.4 6.8
9 Japanese 🟡 7.4 7.6 7.4 6.8
10 German 🟡 7.2 7.6 7.0 7.0
View all 19 languages
Rank Language Overall Accuracy Fluency Style
1 Spanish 🟡 7.8 8.6 7.8 7.8
2 Italian 🟡 7.8 8.0 7.8 7.2
3 Portuguese 🟡 7.6 8.2 7.6 7.2
4 Chinese (Traditional) 🟡 7.6 7.6 7.2 7.2
5 Russian 🟡 7.6 7.8 7.6 7.0
6 Arabic 🟡 7.6 8.0 7.6 7.4
7 French 🟡 7.4 7.4 7.6 7.0
8 Chinese (Simplified) 🟡 7.4 7.6 7.4 6.8
9 Japanese 🟡 7.4 7.6 7.4 6.8
10 German 🟡 7.2 7.6 7.0 7.0
11 Polish 🟡 7.2 7.6 7.4 7.0
12 Ukrainian 🟡 7.0 7.6 6.8 7.0
13 Korean 🟠 6.8 7.6 6.8 6.6
14 Vietnamese 🟠 6.8 7.8 6.8 6.8
15 Hindi 🟠 6.8 7.8 6.8 6.6
16 Thai 🟠 6.6 7.4 6.4 6.6
17 Hebrew 🟠 6.4 7.2 6.2 6.4
18 Tamil 🟠 6.2 7.2 6.2 6.0
19 Bengali 🟠 5.8 6.8 5.8 5.6

Performance by Category

European Major Languages

Language Overall Accuracy Fluency Style
Spanish 🟡 7.8 8.6 7.8 7.8
Italian 🟡 7.8 8.0 7.8 7.2
Portuguese 🟡 7.6 8.2 7.6 7.2
French 🟡 7.4 7.4 7.6 7.0
German 🟡 7.2 7.6 7.0 7.0
Polish 🟡 7.2 7.6 7.4 7.0

Category Average: 🟡 7.5

Asian Languages

Language Overall Accuracy Fluency Style
Chinese (Traditional) 🟡 7.6 7.6 7.2 7.2
Chinese (Simplified) 🟡 7.4 7.6 7.4 6.8
Japanese 🟡 7.4 7.6 7.4 6.8
Korean 🟠 6.8 7.6 6.8 6.6
Vietnamese 🟠 6.8 7.8 6.8 6.8
Hindi 🟠 6.8 7.8 6.8 6.6
Thai 🟠 6.6 7.4 6.4 6.6
Tamil 🟠 6.2 7.2 6.2 6.0
Bengali 🟠 5.8 6.8 5.8 5.6

Category Average: 🟠 6.8

Cyrillic Languages

Language Overall Accuracy Fluency Style
Russian 🟡 7.6 7.8 7.6 7.0
Ukrainian 🟡 7.0 7.6 6.8 7.0

Category Average: 🟡 7.3

Semitic Languages

Language Overall Accuracy Fluency Style
Arabic 🟡 7.6 8.0 7.6 7.4
Hebrew 🟠 6.4 7.2 6.2 6.4

Category Average: 🟡 7.0


Translation Examples

Best Translation

French - Walden Score: 🟢 9.0/10

View Translation

Original:

Translation:

Je suis allé dans les bois parce que je désirais vivre délibérément, affronter uniquement les faits essentiels de la vie et voir si je ne pouvais apprendre ce qu’elle avait à m’enseigner, afin de ne pas découvrir, au moment de mourir, que je n’avais pas vécu. Je ne voulais pas vivre une vie sans substance, car vivre est si précieux ; je ne voulais pas non plus me résigner, à moins que cela ne soit absolument nécessaire. Je voulais vivre intensément et extraire toute la moelle de la vie.

Feedback: The translation faithfully preserves Thoreau's meaning and philosophical intent with excellent accuracy. The French flows naturally and reads as genuine philosophical prose, though 'une vie sans substance' slightly softens the original's 'what was not life,' and 'vivre intensément' is somewhat more conventional than the vivid 'live deep' metaphor.

Worst Translation

Thai - The Picture of Dorian Gray Score: 🟠 5.0/10

View Translation

Original:

Translation:

สตูดิโออบอวลไปด้วยกลิ่นกุหลาบอันหอมหวาน และเมื่อสายลมฤดูร้อนพัดผ่านต้นไม้ในสวน กลิ่นลาเวนเดอร์เข้มข้น หรือกลิ่นหอมละมุนของต้นกุหลาบป่าสีชมพูก็จะลอยเข้ามาตามประตูที่เปิดค้าง ลอร์ดเฮนรี่ ว็อตตันที่นอนอยู่บนโซฟาบุหนังเปอร์เซีย สูบบุหรี่ไม่ขาดสาย สามารถมองเห็นแสงระยิบระยับของดอกลบูนที่สีทองอร่ามราวกับน้ำผึ้งได้จากมุมหนึ่งของโซฟา

Feedback: The translation captures basic meaning but contains significant errors: 'lilac' is mistranslated as 'lavender,' and 'pink-flowering thorn' becomes 'wild rose,' losing Wilde's precise botanical imagery. The sensory richness and aesthetic precision of the original are diminished, and the Thai phrasing feels somewhat awkward and less elegant than the source's flowing prose.


Score Distribution

Score Range Count Percentage
🟢 Excellent (9-10) 1 1.1%
🟡 Good (7-8) 73 76.8%
🟠 Acceptable (5-6) 21 22.1%
🔴 Poor (3-4) 0 0.0%
⚫ Failed (1-2) 0 0.0%

Performance Metrics

Metric Value
Average Translation Time 3384.0ms
Success Rate 100.0%

← Back to Home | All Models

Clone this wiki locally