Benchmarks · composite
Arena Text · Korean
Crowdsourced pairwise human preference on open-ended text prompts (style control on); Bradley-Terry rating. Only the Korean category.
171 models · 182 results · data as of 2026-10-29 · arena.ai ↗
Leaderboard
Fable 5 high
All results
182 rows| # | Model | Effort | Arena Text · Korean |
|---|---|---|---|
| 1 | Fable 5 | high | 1497 ±22 |
| 2 | Opus 5 | max | 1495 ±24 |
| 3 | Fable 5.1 | max | 1494 ±37 |
| 4 | Opus 5 | high | 1493 ±18 |
| 5 | Muse Spark 1.3 | max | 1491 ±31 |
| 6 | Gemini 3.7 Flash | high | 1481 ±26 |
| 7 | Muse Spark | 1474 ±39 | |
| 8 | Muse Spark 1.1 | max | 1474 ±21 |
| 9 | Opus 4.7 | max | 1471 ±19 |
| 10 | Kimi K3 | max | 1470 ±22 |
| 11 | GPT-6.1 Sol | max | 1467 ±44 |
| 12 | Gemini 3.8 Flash | high | 1465 ±23 |
| 13 | DeepSeek V4 Pro | max | 1464 ±40 |
| 14 | Muse Spark 1.2 | xhigh | 1463 ±37 |
| 15 | Opus 4.6 | max | 1463 ±17 |
| 16 | Opus 4.7 | high | 1461 ±19 |
| 17 | GLM 5.3 | max | 1460 ±28 |
| 18 | Qwen 3.8 Max | max | 1456 ±24 |
| 19 | Gemini 3.1 Pro | high | 1455 ±14 |
| 20 | GPT-5.6 Sol | xhigh | 1455 ±22 |
| 21 | GPT-6 Astra | max | 1453 ±38 |
| 22 | GPT-5.5 Instant | 1451 ±30 | |
| 23 | Gemini 3 Pro | high | 1449 ±23 |
| 24 | Opus 4.5 | high | 1448 ±18 |
| 25 | Opus 4.8 | high | 1448 ±19 |
| 26 | Opus 4.6 | high | 1447 ±17 |
| 27 | Gemini 3 Flash | high | 1445 ±25 |
| 28 | DeepSeek V4.1 Flash | max | 1444 ±38 |
| 29 | GPT-5.5 | high | 1444 ±18 |
| 30 | Grok 4.20 | xhigh | 1444 ±28 |