Up to four models
Llama 2-13B vs Mistral 7B v0.3
On the index (GPT-4 = 100), Llama 2-13B scores 87 and Mistral 7B v0.3 82. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
Llama 2-13B and Mistral 7B v0.3 are too close to call (6 points apart, within the uncertainty).
Test by testthird-party results
Every result as a table
| Test | Llama 2-13B | Mistral 7B v0.3 |
|---|---|---|
| Reasoning & knowledge | ||
| MMLU (HELM runs) | 55.4% | 59.9% |
| Agents | ||
| DTBench | 42.2% | 42.5% |
Tests at least two of these models have taken. Bold rows are the core tests.