Modellvergleich
Wählen Sie bis zu 4 Modelle aus und vergleichen Sie sie über alle Metriken.
Beliebte Duelle:
| Metrik | Claude Mythos Preview | Claude 5.5 Sonnet | Kimi K3.1 Pro |
|---|---|---|---|
| Anbieter | Anthropic | Anthropic | Moonshot AI |
| Lizenz | Proprietary | Proprietary | Open Weights |
| Podium Score | 97.4 | 94.8 | 94.2 |
| Gesamtrang | 1 | 2 | 3 |
| Arena Elo | — | 1520 | 1490 |
| Intelligenz-Index | — | 95.0 | 68.0 |
| Coding | 80.9 | 91.5 | 72.0 |
| Mathematik | 100.0 | 94.2 | 95.0 |
| Reasoning | 98.8 | 95.0 | 98.0 |
| Agentic | — | 89.4 | 82.0 |
| Wissen | 100.0 | 88.0 | 45.0 |
| Multimodal | 100.0 | 96.0 | 96.0 |
| Langkontext | — | 98.0 | 100.0 |
| Tempo | — | 78.0 | 25.0 |
| Preis-Leistung | — | 82.0 | 55.0 |
| Ausgabegeschwindigkeit | 80 t/s | 94 t/s | 45 t/s |
| Eingabepreis ($/M) | $3/M | $3/M | $4/M |
| Ausgabepreis ($/M) | $15/M | $15/M | $20/M |
| Kontextfenster | 1000K | 1000K | 1000K |
| SWE-Bench Verified | 93.9% | 91.5% | 91.5% |
| GPQA Diamond | 94.6% | 95% | 97.5% |
| Multilingual MMLU | 92.7% | — | — |
| Browsecomp | 86.9% | 95.2% | 94% |
| CharXiv Reasoning | 93.2% | — | — |
| Humanity’s Last Exam | 64.7% | 61.2% | 58% |
| SWE-Bench Pro | 77.8% | 74.8% | 70% |
| FrontierMath | — | 94.2% | 95% |
| OSWorld | — | 84.5% | — |
| LiveCodeBench | — | 84% | 80% |
| MMMU-Pro | — | 91% | 88% |
| MRCR v2 | — | 96.4% | 96% |
| τ²-Bench Retail | — | 92.5% | 90% |
| Terminal-Bench Hard | — | 88% | 76% |
| Ifbench | — | 89.5% | 80% |