math Math · llm-stats.com
FrontierMath
Research-grade mathematics problems crafted by professional mathematicians.
7
Getestete Modelle
GPT-6 Astra
Top-Modell
97.6%
Top-Score
OpenAI
Anbieter
| # | Modell | FrontierMath |
|---|---|---|
| 1 | GPT-6 Astra OpenAI | 97.6% |
| 2 | GPT-5.6 Sol OpenAI | 89% |
| 3 | GPT-5.6 Terra OpenAI | 84.9% |
| 4 | GPT-5.6 Luna OpenAI | 78.6% |
| 5 | GPT-5.4 OpenAI | 47.6% |
| 6 | GPT-5.2 OpenAI | 40.3% |
| 7 | GPT-5.5 OpenAI | 35.4% |
Benchmark-Daten aggregiert aus llm-stats.com. Siehe Methodik.