Modell-Rankings
Modell-Rankings
Jede Quelle hat eigene Regeln — kein Gesamt-#1. Ausfälle: Status-Tab.
Arena Elo
People pick the better answer in blind A/B chats; that becomes Elo.
Schaut auf: How chat and writing feel to people.
Schaut nicht auf: Truthfulness, security, price, speed, or coding skill.
Top
Elo →
- 01Claude Fable 5anthropic1509 Elo
- 02Claude Opus 4 6 Thinkinganthropic1505 Elo
- 03Claude Opus 4 7 Thinkinganthropic1502 Elo
- 04Claude Opus 4 6anthropic1497 Elo
- 05Qwen3.8 Maxalibaba1496 Elo
- 06Claude Opus 4 7anthropic1492 Elo
- 07Claude Opus 5Highanthropic1492 Elo
- 08Claude Opus 5 Maxanthropic1490 Elo
- 09Muse Spark 1.1meta1490 Elo
- 10Muse Sparkmeta1488 Elo
- 11Gemini 3 Progoogle1486 Elo
- 12Gemini 3.1 Pro Previewgoogle1485 Elo
- 13Kimi K3 Maxmoonshot1485 Elo
- 14Claude Opus 4 8 Thinkinganthropic1484 Elo
- 15GPT 5.6 SolXHIGHopenai1483 Elo
- 16Gemini 3.6 Flashgoogle1483 Elo
- 17GPT 5.5Highopenai1482 Elo
- 18GPT 5.4Highopenai1477 Elo
- 19Gemini 3.5 FlashHighgoogle1476 Elo
- 20GPT 5.5openai1476 Elo
Bars are scaled to make gaps easier to see.
Gleiches Modell, andere Quellen. Leer = nicht in der Top-Liste.