Benchmarks · math
GSM8K
Uses Epoch AI's external GSM8K exact-match leaderboard. Score: GSM8K exact match. Collected by Jake Boggs' benchmark dashboard.
18 models · 18 results · data as of 2026-10-29 · epoch.ai ↗
Leaderboard
Qwen 2.5 Coder none
All results
18 rows| # | Model | Effort | GSM8K |
|---|---|---|---|
| 1 | Qwen 2.5 Coder | none | 93.0% |
| 2 | GPT-4 | high | 92.0% |
| 3 | GPT-4o mini | high | 91.3% |
| 4 | Gemma 2 9B | 84.9% | |
| 5 | Mistral | none | 84.2% |
| 6 | Gemini 1.5 Flash | none | 82.4% |
| 7 | Llama 3.1 8b | medium | 82.4% |
| 8 | GPT-3.5 | high | 78.2% |
| 9 | Mixtral 8x7B Instruct | none | 74.4% |
| 10 | Llama 2 70b | none | 69.6% |
| 11 | Llama 6 5b | 54.4% | |
| 12 | Llama 3 3b | 44.1% | |
| 13 | Llama 2 34b | 42.2% | |
| 14 | Llama 2 13b | none | 36.9% |
| 15 | Qwen 1 | 21.2% | |
| 16 | Llama 1 3b | 20.5% | |
| 17 | Llama 2 7b | none | 16.7% |
| 18 | GPT-3 | medium | 9.0% |