Google: Gemma 4 E4B
1 benchmarks
Leaderboard position
#1008
Overall rank (of 1121 models)
-124
Elo rating
1
Benchmarks
Top 89.9%
Percentile
In Google (of 72 models)
Above: Apple: openelm-450m #1007
Below: Qwen: qwen2.5-vl-7b-instruct #1009
Benchmark highlights
Best benchmark
WikiText-2 (-ppl)
-23.1498
Worst benchmark
WikiText-2 (-ppl)
-23.1498
Nearest competitor from another company
The model from another company with the most similar benchmark scores.
appeared as nearest in 1 of 1 benchmarks
Benchmarks
Predicted scores
Estimated from 25 similar models and the global Elo rating. These are projections, not measured results.
| Benchmark | Predicted | Confidence | Based on |
|---|---|---|---|
| MATH-500 | ~69.41 | 50% | — |
| AIME 2025 | ~34 | 50% | — |
| AIME 2026 | ~54.57 | 50% | — |
| HMMT Feb 2026 | ~46.01 | 50% | — |
| GPQA Diamond | ~36.56 | 50% | — |
| MMLU-Pro | ~37.22 | 50% | — |
| Humanity's Last Exam (no tools) | ~4 | 50% | — |
| IFEval | ~68.91 | 50% | — |
| IFBench | ~37.14 | 50% | — |
| HumanEval | ~45.3 | 50% | — |
| MBPP | ~39.82 | 50% | — |
| Terminal-Bench v2.1 | ~8.6 | 50% | — |
| MMLU | ~42.46 | 50% | — |
| SWE-bench Verified | ~8.5 | 50% | — |
| SWE-bench Pro | ~5.24 | 50% | — |
| SciCode | ~11.64 | 50% | — |
| AA-LCR | ~23.98 | 50% | — |
| Artificial Analysis Intelligence Index v5.0.1 | ~12 | 50% | — |
| Humanity's Last Exam (with tools) | ~11.9 | 50% | — |
| GSM8K | ~46.02 | 50% | — |
| -(CoderPPL) | ~-6.98 | 50% | — |
| LiveCodeBench v6 | ~33.58 | 50% | — |
| MMMU-Pro | ~34.58 | 50% | — |
| MATH | ~19.7 | 50% | — |
| BBH | ~43.62 | 50% | — |