Language Understanding on MMLU (Accuracy, Δ (pts), and padj)
6.2MMLU Δ (pts)Mixtral 8x22B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Mixtral 8x22Bshots=5-shot, comparison_baseline=GPT-3.5 Turbo2026.07 | 6.2 | — | 0.0036 | |
| Nemotron-4 340Bshots=5-shot, comparison_baseline=Mixtral 8x22B2026.07 | 5.7 | — | 0.0036 | |
| Llama 3 405Bshots=5-shot, comparison_baseline=GPT-4 (0125)2026.07 | 2.2 | — | 0.0036 | |
| GPT-4oshots=5-shot, comparison_baseline=Llama 3 405B2026.07 | 1.8 | — | 0.0036 | |
| GPT-4 (0125)shots=5-shot, comparison_baseline=Llama 3 70B2026.07 | 1.5 | — | 0.0036 | |
| GPT-3.5 Turboshots=5-shot, comparison_baseline=Llama 3 8B2026.07 | 1.3 | — | 0.065 | |
| Llama 3 70Bshots=5-shot, comparison_baseline=Nemotron-4 340B2026.07 | 1 | — | 0.065 | |
| Claude 3.5 Sonnetshots=5-shot, comparison_baseline=GPT-4o2026.07 | 0.8 | — | 0.065 |