DataLearner logo

DeepSeek-V3 Benchmark Details

DeepSeek-V3 currently shows benchmark results led by BBH (3 / 21, score 92.30), MATH (7 / 42, score 87.80), HumanEval (9 / 39, score 89).

Benchmark Results

DeepSeek-V3

Benchmark Results

Thinking

General Knowledge

5 evaluations
Benchmark / mode
Score
Rank/total
BBH
92.30
3 / 21
MMLU
88.50
17 / 66
MMLU Pro
75.90
83 / 132
GPQA Diamond
59.10
148 / 187
GPQA
59.10
6 / 15

Coding and Software Engineer

2 evaluations
Benchmark / mode
Score
Rank/total
HumanEval
89
9 / 39
LiveCodeBench
34.60
110 / 123

Math and Reasoning

4 evaluations
Benchmark / mode
Score
Rank/total
MATH
87.80
7 / 42
MATH-500
87.80
39 / 44
AIME 2024
39
52 / 62
FrontierMath
1.70
49 / 60

Common Sense

1 evaluations
Benchmark / mode
Score
Rank/total
SimpleQA
24.90
31 / 47

Writing and Creative Capabilities

1 evaluations
Benchmark / mode
Score
Rank/total
Creative Writing
81.60
15 / 23

Common Sense Reasoning

1 evaluations
Benchmark / mode
Score
Rank/total
Simple Bench
Standard Mode
18.90
59 / 63

Agent Level Benchmark

1 evaluations
Benchmark / mode
Score
Rank/total
Aider-Polyglot
Standard Mode
48.40
34 / 59
Compare with other models