EffiBench - NMU — leaderboard
Metric: Normalized Memory Usage. Source: huggingface.co. 24 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | starcoder2-15B | 1.71 |
| 2 | CodeLlama-13B-hf | 1.85 |
| 3 | Yi 34B (Chat) | 1.89 |
| 4 | Yi 34B (Base) | 1.89 |
| 5 | deepseek-coder-6.7B-instruct | 1.9 |
| 6 | Claude 3 Haiku | 1.91 |
| 7 | Claude 3 Sonnet | 1.91 |
| 8 | starcoder | 1.91 |
| 9 | GPT-3.5 Turbo (0301) | 1.91 |
| 10 | deepseek-coder-1.3B-instruct | 1.91 |
| 11 | deepseek-coder-6.7B-base | 1.91 |
| 12 | GPT-4 | 1.92 |
| 13 | CodeLlama-34B-hf | 1.92 |
| 14 | CodeLlama-13B-Instruct-hf | 1.92 |
| 15 | GPT-3.5 Turbo (0613) | 1.92 |
Interactive version: aibenchmarks.dev/benchmark?slug=effibench-nmu · How the rankings work · Data refreshed daily, snapshot 2026-07-17.