Benchmarks · reasoning
AlpacaEval 2.0 Verified
Uses Tatsu Lab's verified/minimal AlpacaEval 2.0 length-controlled win-rate leaderboard. Score: length-controlled win rate. Collected by Jake Boggs' benchmark dashboard.
20 models · 20 results · data as of 2026-10-29 · tatsu-lab.github.io ↗
Leaderboard
GPT-4o high
All results
20 rows| # | Model | Effort | AlpacaEval 2.0 Verified |
|---|---|---|---|
| 1 | GPT-4o | high | 57.5% |
| 2 | GPT-4 | high | 55.0% |
| 3 | GPT-4o mini | high | 50.7% |
| 4 | Opus 02 | 40.5% | |
| 5 | Llama 3.1 405b | none | 39.3% |
| 6 | Qwen 2 72B | none | 38.1% |
| 7 | Llama 3.1 70b | none | 38.1% |
| 8 | Qwen 1.5 72B | 36.6% | |
| 9 | Sonnet 02 | 34.9% | |
| 10 | Llama 3 70b | none | 34.4% |
| 11 | Mistral Large 24 | 32.7% | |
| 12 | Mistral | none | 29.7% |
| 13 | Mistral Medium | none | 28.6% |
| 14 | Qwen 1.5 14B | 23.9% | |
| 15 | Llama 3 8b | none | 22.9% |
| 16 | GPT-3.5 | high | 22.7% |
| 17 | Llama 3.1 8b | medium | 20.9% |
| 18 | Qwen 1.5 7B | 14.7% | |
| 19 | Llama 3 3b | 9.9% | |
| 20 | Qwen 1.5 1.8B | 2.6% |