Verdict
DeepSeek-R1 leads 6–1 across 7 shared benchmarks.
DeepSeek-R1 6 · DeepSeek-R1-Distill-Qwen-32B 1 · higher isn't always better — check the caveats below.
Reported scores — protocols may differ. How we compare
Math: DeepSeek-R1 leads 2–1 across 3 shared math benchmarks (largest gap: AIME, 79.8 vs 72.6 — single evaluator).
Coding: DeepSeek-R1 leads 2–0 across 2 shared coding benchmarks (largest gap: Codeforces rating, 2029 vs 1691 — single evaluator).
Vendor claims — compare against the measured scores above. A claim is what a developer says about its own model, not an independent measurement.
Reasoning
“After these steps, we obtained a checkpoint referred to as DeepSeek-R1, which achieves performance on par with OpenAI-o1-1217.”
“For education-oriented knowledge benchmarks such as MMLU, MMLU-Pro, and GPQA Diamond, DeepSeek-R1 demonstrates superior performance compared to DeepSeek-V3.”
Reasoning
“Using Qwen2.5-32B as the base model, direct distillation from DeepSeek-R1 outperforms applying RL on it.”
DeepSeek-R1 leads 6–1 across 7 shared benchmarks. DeepSeek-R1 leads 6 benchmarks and DeepSeek-R1-Distill-Qwen-32B leads 1. Higher isn't always better — see the integrity caveats on each benchmark.
DeepSeek-R1 and DeepSeek-R1-Distill-Qwen-32B have 7 benchmarks in common in our data — those are the only rows where a direct, apples-to-apples comparison is drawn.
DeepSeek-R1 leads coding 2–0. DeepSeek-R1 leads math 2–1.