benchgap
DeepSeek · model

DeepSeek R1 Distill Llama 70B benchmark scores

As of 2026-11-19, DeepSeek R1 Distill Llama 70B (DeepSeek) has measured scores on 4 benchmarks and estimated scores on 20 more.

BenchmarkScoreSource
AA-GPQA Diamond63.1%estimated ± 6.2 pp, medium confidence
AA-HLE6.5%estimated ± 5.7 pp, medium confidence
Artificial Analysis Intelligence Index7.9%estimated ± 7.4 pp, medium confidence
CritPt0.3%estimated ± 4.0 pp, high confidence
GPQA Diamond65.5%estimated ± 1.2 pp, high confidence
HLE6.8%estimated ± 6.5 pp, medium confidence
AA-Omniscience Accuracy19.1%estimated ± 8.0 pp, medium confidence
SWE-bench Verified56.6%estimated ± 9.3 pp, low confidence
MMLU-Pro78.9%estimated ± 4.4 pp, high confidence
AIME 202470.0%measured
GPQA65.2%measured
MMLU87.8%estimated ± 4.2 pp, high confidence
Vals GPQA Diamond63.1%estimated ± 4.5 pp, high confidence
Vals MMLU-Pro76.7%estimated ± 2.3 pp, high confidence
MATH-50094.5%measured
MMMLU80.9%estimated ± 1.7 pp, high confidence
AIME 202556.5%estimated ± 3.5 pp, high confidence
AA MMLU-Pro84.8%estimated ± 1.2 pp, low confidence
HLE w/o tools11.6%estimated ± 5.1 pp, low confidence
C-Eval87.2%estimated ± 1.2 pp, high confidence
MMLU-Redux89.2%estimated ± 1.7 pp, high confidence
SuperGPQA45.1%estimated ± 8.5 pp, low confidence
SimpleQA24.8%estimated ± 13.4 pp, low confidence
LiveCodeBench57.5%measured