Mathematical Reasoning on MGSM Rev2
99.4Overall Result AccuracyGEMINI-3-FLASH
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| GEMINI-3-FLASHResource Profile=Low2026.06 | 99.4 | — | — | — | — | — | — | — | 0.5 | 99.5 | |
| GEMINI-3-FLASHResource Profile=Standard2026.06 | 99.3 | — | — | — | — | — | — | — | 0.4 | 99.6 | |
| CLAUDE-SONNET-4.6Model Version=4.62026.06 | 98.7 | — | — | — | — | — | — | — | -2.6 | 99.4 | |
| GEMMA-4-31B-ITModel Variant=31B-IT2026.06 | 98.3 | — | — | — | — | — | — | — | 1.4 | 98.4 | |
| GEMINI-2.5-FLASHThinking=On2026.06 | 98.2 | — | — | — | — | — | — | — | 1.3 | 98.7 | |
| GPT-OSS-120BParameter Count=120B2026.06 | 97 | — | — | — | — | — | — | — | 2.6 | 97.4 | |
| GEMMA-4-26B-A4B-ITModel Variant=26B-A4B-IT2026.06 | 96.5 | — | — | — | — | — | — | — | 3 | 96.8 | |
| GPT-OSS-20BParameter Count=20B2026.06 | 95.2 | — | — | — | — | — | — | — | 3.8 | 96.1 | |
| GEMMA-3-27BParameter Count=27B2026.06 | 93.3 | — | — | — | — | — | — | — | 4.1 | 95.3 | |
| CLAUDE-OPUS-4.7Model Version=4.72026.06 | 90.4 | — | — | — | — | — | — | — | -1 | 97.3 | |
| QWEN-3-30B-A3BModel Variant=30B-A3B2026.06 | 88.6 | — | — | — | — | — | — | — | 9.8 | 89.8 | |
| GEMMA-3-12BParameter Count=12B2026.06 | 88.1 | — | — | — | — | — | — | — | 7.9 | 91.3 | |
| GEMINI-3-FLASHResource Profile=Minimal2026.06 | 86.5 | — | — | — | — | — | — | — | 0.4 | 96.2 | |
| QWEN-3-4BParameter Count=4B2026.06 | 79.4 | — | — | — | — | — | — | — | 18.1 | 81.3 | |
| QWEN3 4Bn=42026.04 | 75.5 | 74.7 | 75.5 | 75.5 | 73.4 | 75.6 | 75.9 | 73.7 | — | — | |
| QWEN3 4Bn=82026.04 | 75.4 | 74.8 | 75.6 | 75.6 | 73.5 | 74.6 | 74.6 | 73.2 | — | — | |
| QWEN3 4Bn=162026.04 | 75.4 | 74 | 75.5 | 75.8 | 73.2 | 74.8 | 74.7 | 73 | — | — | |
| QWEN3 8Bn=322026.04 | 74.7 | 73.8 | 75.6 | 75.1 | 72.6 | 75.6 | 74.1 | 71.7 | — | — | |
| QWEN3 4Bn=322026.04 | 74.6 | 74.5 | 75.3 | 75.3 | 72.3 | 76.1 | 74.3 | 72.7 | — | — | |
| QWEN3 8Bn=162026.04 | 74.4 | 73.8 | 74.1 | 74.3 | 73.3 | 74.9 | 73.9 | 72.1 | — | — | |
| QWEN3 8Bn=82026.04 | 74.1 | 73.2 | 74.3 | 74 | 73.8 | 74 | 73.8 | 72.8 | — | — | |
| QWEN3 8Bn=42026.04 | 73.9 | 73.3 | 73.9 | 74.6 | 73.5 | 74.1 | 73.6 | 72.1 | — | — | |
| GEMMA-3-4BParameter Count=4B2026.06 | 72.9 | — | — | — | — | — | — | — | 13.4 | 82.6 | |
| DISTILL-QWEN 8Bn=162026.04 | 62.6 | 62.5 | 61.5 | 63.2 | 63.2 | 63.9 | 60 | 60.7 | — | — | |
| DISTILL-QWEN 8Bn=42026.04 | 62.3 | 62.7 | 62.3 | 63 | 62.2 | 63.3 | 60.4 | 63.3 | — | — | |
| DISTILL-QWEN 8Bn=322026.04 | 61.5 | 62.3 | 58.9 | 64 | 61.7 | 65.1 | 60.9 | 59.6 | — | — | |
| DISTILL-QWEN 8Bn=82026.04 | 61.2 | 62.3 | 62.2 | 62.6 | 62.9 | 64.1 | 61.2 | 63.6 | — | — | |
| GEMINI-2.5-FLASHThinking=Off2026.06 | 55.6 | — | — | — | — | — | — | — | 2.8 | 89.8 | |
| CLAUDE-HAIKU-4.5Model Version=4.52026.06 | 54 | — | — | — | — | — | — | — | -1.3 | 86 | |
| DISTILL-QWEN 1.5Bn=322026.04 | 36.8 | 36.9 | 37.1 | 43 | 36 | 40.2 | 39.1 | 34.3 | — | — | |
| DISTILL-QWEN 1.5Bn=162026.04 | 36.1 | 36.5 | 36.5 | 40.2 | 36.1 | 39.6 | 38.7 | 35.1 | — | — | |
| DISTILL-QWEN 1.5Bn=42026.04 | 36 | 36.6 | 36.7 | 37.8 | 36.7 | 38.1 | 38.1 | 36 | — | — | |
| DISTILL-QWEN 1.5Bn=82026.04 | 36 | 36.8 | 36.7 | 39.5 | 36.7 | 38.7 | 38.3 | 34.9 | — | — | |
| GEMMA-3-1BParameter Count=1B2026.06 | 25.9 | — | — | — | — | — | — | — | 24.7 | 47.7 | |
| GEMMA-4-E4B-ITModel Variant=E4B-IT2026.06 | 17.5 | — | — | — | — | — | — | — | 4.4 | 68 | |
| GEMMA-4-E2B-ITModel Variant=E2B-IT2026.06 | 13 | — | — | — | — | — | — | — | 5 | 53.7 |