DataLearner logo

Qwen3-235B-A22B Benchmark Details

Qwen3-235B-A22B currently shows benchmark results led by GSM8K (2 / 70, score 96.40), MATH-500 (7 / 45, score 98), MMLU (38 / 124, score 85.80).

Benchmark Results

Qwen3-235B-A22B

Benchmark Results

Thinking
Tool usage

General Knowledge

7 evaluations
Benchmark / mode
Score
Rank/total
BBH
Standard Mode
88.87
8 / 21
MMLU
Standard Mode
85.80
38 / 124
MMLU Pro
Standard Mode
72.90
93 / 176
HLE
Standard Mode
7.60
427 / 563
HLE
Standard Mode
4.20
513 / 563
HLE
Thinking Mode
11
377 / 563
ARC-AGI-1
Standard Mode
4.30
145 / 147

Math and Reasoning

7 evaluations
Benchmark / mode
Score
Rank/total
MATH-500
Standard Mode
96.20
19 / 45
MATH-500
Thinking Mode
98
7 / 45
GSM8K
Standard Mode
96.40
2 / 70
AIME 2024
Standard Mode
85.70
20 / 62
AIME 2024
Thinking Mode
85.70
20 / 62
AIME2025
Standard Mode
24.70
193 / 215
AIME2025
Thinking Mode
81.50
92 / 215

Reading Comprehension

1 evaluations
Benchmark / mode
Score
Rank/total
DROP
Standard Mode
88.70
5 / 9

Other

2 evaluations
Benchmark / mode
Score
Rank/total
GPQA Diamond
Standard Mode
71.10
314 / 462
GPQA Diamond
Thinking Mode
71.10
314 / 462

Common Sense

1 evaluations
Benchmark / mode
Score
Rank/total
SimpleQA
Standard Mode
11
40 / 47

Coding and Software Engineer

3 evaluations
Benchmark / mode
Score
Rank/total
LiveCodeBench
Standard Mode
70.70
86 / 250
LiveCodeBench
Thinking Mode
70.70
86 / 250
SWE-bench Verified
Standard Mode
34.40
109 / 116

Common Sense Reasoning

1 evaluations
Benchmark / mode
Score
Rank/total
SimpleBench
Thinking Mode
31
74 / 93

Agent Level Benchmark

6 evaluations
Benchmark / mode
Score
Rank/total
Aider-Polyglot
Standard Mode
59.60
23 / 59
τ²-Bench
Thinking ModeTools
34.40
42 / 44
τ²-Bench - Telecom
Standard ModeTools
27.20
222 / 264
τ²-Bench - Telecom
Thinking ModeTools
24
233 / 264
Terminal Bench Hard
Standard ModeTools
6.10
204 / 244
Terminal Bench Hard
Thinking ModeTools
6.10
204 / 244

Instruction Following

2 evaluations
Benchmark / mode
Score
Rank/total
IF Bench
Standard Mode
36.60
237 / 282
IF Bench
Thinking Mode
38.70
227 / 282
Compare with other models