AI Energy Score (Reasoning): leaderboard
Hugging Face energy efficiency ratings measuring power consumption of reasoning-enabled AI models, which use up to 30x more energy than non-reasoning equivalents.
Metric: Energy Score (1-5). Source: huggingface.co. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 Next 80B A3B (Thinking) | 5 | #320 (Qwen 3 Next 80B A3B) |
| 2 | Qwen 3 Next 80B A3B Instruct | 5 | #368 |
| 3 | Qwen 3 235B A22B | 5 | #313 |
| 4 | Qwen 3 0.6B | 5 | #1459 |
| 5 | GPT-OSS-20B (Low) | 5 | #518 (GPT-OSS-20B) |
| 6 | GPT-OSS-20B (Medium) | 5 | #518 (GPT-OSS-20B) |
| 7 | GPT-OSS-20B (High) | 4 | #518 (GPT-OSS-20B) |
| 8 | Qwen 3 30B A3B | 3 | #493 |
| 9 | DeepSeek R1 Distill Qwen 32B | 3 | #655 |
| 10 | Phi-4-reasoning-plus (Thinking) | 2 | |
| 11 | GPT-OSS-120B (Low) | 2 | #341 (GPT-OSS-120B) |
| 12 | GPT-OSS-120B (Medium) | 2 | #341 (GPT-OSS-120B) |
| 13 | gemma-3-4B-pt Base | 1 | #1134 |
| 14 | DeepSeek R1 Distill Llama 70B | 1 | #594 |
| 15 | GPT-OSS-120B (High) | 1 | #341 (GPT-OSS-120B) |
Interactive version: theaggregate.ai/benchmark?slug=ai-energy-score-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-20.