DeepSeek R1 Zero: benchmark results
Provider: DeepSeek. Released 2025-01-20. Access: Open.
Aggregate Rating 1663 ± 27, rank #536 of 10916 rated models, from 12 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SpeechMap Compliance | 91.6 | % Requests Completed | 93.3 |
| BenchTable - STEM | 80.9 | Weighted Score (%) | 93 |
| BenchTable | 70.5 | Total Score (%) | 87.1 |
| SuperGPQA | 60.24 | Accuracy (%) | 86 |
| BenchTable - Reasoning | 66.7 | Weighted Score (%) | 81.8 |
| LLM Stats (AIME 2024) | 86.7 | Score (%) | 75 |
| BenchTable - Tech | 61.8 | Weighted Score (%) | 60.2 |
| BenchTable - Utility | 53.6 | Weighted Score (%) | 50.9 |
| ZeroEval GPQA Diamond | 73.3 | GPQA Diamond Score | 47.3 |
| ZeroEval MATH-500 | 95.9 | MATH-500 Score | 46.9 |
| LLM Stats Score | 15.45 | LLM Stats Score (conservative rating) | 34.9 |
| Reward Hacking Benchmark (RHB) | 13.9 | Exploit rate (%) (self-reported) | 0 |
Interactive version: theaggregate.ai/model?slug=deepseek-r1-zero · How It Works · Data refreshed daily, snapshot 2026-11-08.