AI BENCHY Kategoriefehler
Anti-KI-Tricks: Falsche Antwort
Anti-KI-Tricks
Falsche Antwort
Sieh, welche KI-Modelle bei Anti-KI-Tricks am ehesten auf Falsche Antwort stoßen, damit du Schwachstellen schneller erkennst. Sortieren nach: Fehleranzahl ↑.
Fehlergründe
| Rang | Modell | Unternehmen | Falsche Antwort-Anzahl | Kategorie-Score | Korrekte Tests | Antwortzeit (Durchschnitt) |
|---|---|---|---|---|---|---|
| #3 | Claude Opus 4.7 medium | Anthropic | 1 | 8.3 | 3/4 | 1.85s |
| #4 | Claude Opus 4.7 none | Anthropic | 1 | 8.3 | 3/4 | 2.12s |
| #6 | Seed-2.0-Lite medium | Bytedance Seed | 1 | 8.3 | 3/4 | 18.0s |
| #7 | GPT-5.3-Codex medium | OpenAI | 1 | 8.7 | 3/4 | 4.16s |
| #8 | Qwen3.5 Plus 2026-02-15 medium | Qwen | 1 | 8.2 | 3/4 | 45.8s |
| #15 | Gemini 2.5 Flash medium | 1 | 8.4 | 3/4 | 6.30s | |
| #16 | GPT-5.4 medium | OpenAI | 1 | 8.3 | 3/4 | 4.11s |
| #21 | Gemini 3 Flash Preview none | 1 | 8.3 | 3/4 | 1.25s | |
| #22 | Gemini 3.1 Flash Lite Preview low | 1 | 8.3 | 3/4 | 2.12s | |
| #25 | Grok 4.20 Beta medium | X AI | 1 | 8.7 | 3/4 | 3.16s |
| #26 | Claude Sonnet 4.6 medium | Anthropic | 1 | 6.5 | 2/4 | 2.98s |
| #27 | DeepSeek V3.2 medium | DeepSeek | 1 | 8.4 | 3/4 | 30.7s |
| #28 | GPT-5.2 Chat none | OpenAI | 1 | 8.7 | 3/4 | 3.40s |
| #29 | Gemini 3.1 Flash Lite Preview none | 1 | 7.5 | 2/4 | 1.04s | |
| #31 | GLM 5V Turbo medium | Z.ai | 1 | 7.2 | 2/4 | 10.8s |