Reasoning-Benchmarks

Komposit-Index aus Reasoning- und Wissens-Benchmarks, u. a. GPQA, HLE, MMLU-Pro

Stand der Quelle: 13. August 2026 Bei Beirat validiert: 16. August 2026 Quelle: Artificial Analysis Index (höher = besser) Top 25 von 599 Modellen
# Modell Anbieter AA Intelligence Index
1 Claude Opus 5Adaptive Reasoning, Max Effort Anthropic 63.10
2 Claude Opus 5Adaptive Reasoning, Xhigh Effort Anthropic 62.50
3 Claude Fable 5Adaptive Reasoning, Max Effort, Opus 4.8 Fallback Anthropic 62.10
4 Claude Opus 5Adaptive Reasoning, High Effort Anthropic 61.50
6 Grok 4.6high SpaceXAI 60.90
5 GPT-5.6 Solmax OpenAI 60.90
7 Kimi K3max Kimi 59.70
8 GPT-5.6 Solxhigh OpenAI 59.00
9 Claude Opus 5Adaptive Reasoning, Medium Effort Anthropic 58.60
10 Qwen3.8 Max Alibaba 58.10
11 Qwen3.8 2.4T A95B Alibaba 57.70
13 Claude Opus 4.8Adaptive Reasoning, Max Effort Anthropic 57.30
12 GPT-5.6 Solhigh OpenAI 57.30
7 Kimi K3 Kimi 57.10
14 Muse Spark 1.2xhigh Meta 56.80
15 GPT-5.6 Terramax OpenAI 56.60
16 GPT-5.5xhigh OpenAI 56.30
17 Gemini 3.7 Flashhigh Google 56.00
18 Grok 4.5high SpaceXAI 55.80
19 GPT-5.6 Solmedium OpenAI 55.60
20 Claude Sonnet 5Adaptive Reasoning, Max Effort Anthropic 55.30
21 Claude Opus 4.7Adaptive Reasoning, Max Effort Anthropic 55.00
22 GPT-5.5high OpenAI 54.70
23 Gemini 3.7 Flashmedium Google 53.40
25 Muse Spark 1.1xhigh Meta 53.20

Zeigt die Top 25 von 599 getesteten Modellen. Die vollständige Liste steht bei der Originalquelle.

Methodik & Datenquellen →