Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Software Engineering Issue Solving on SWE-Bench Verified
Loading...
54
Accuracy
OpenThinkerAgent-32B
4.704
17.502
30.3
43.098
Dec 1, 2025
Jan 4, 2026
Feb 7, 2026
Mar 13, 2026
Apr 16, 2026
May 20, 2026
Jun 23, 2026
Accuracy
Relative Gains
Updated 1mo ago
Evaluation Results
Method
Method
Links
Accuracy
Relative Gains
OpenThinkerAgent-32B
Train Size=100K, Metho...
2026.06
54
-
SWE-Lego-Qwen3-32B
Train Size=18K, Method...
2026.06
51
-
SERA-32B
Train Size=25K, Method...
2026.06
49.4
-
PromptBridge
Model Role=Target Mode...
2025.12
46
27.39
Optimized
Model Role=Source Mode...
2025.12
42.2
-
DeepSWE-Preview
Train Size=4.5K, Metho...
2026.06
42.2
-
Nemotron-Terminal-32B
Train Size=264K, Metho...
2026.06
41.9
-
SA-SWE-32B
Train Size=4.5K, Metho...
2026.06
39.4
-
Original
Model Role=Source Mode...
2025.12
38.6
-
Direct Transfer
Model Role=Target Mode...
2025.12
33.4
0
Original
Model Role=Target Mode...
2025.12
32
-
Qwen3-32B
Train Size=N/A, Method...
2026.06
29.1
-
PromptBridge
Model Role=Target Mode...
2025.12
8.8
15.79
Direct Transfer
Model Role=Target Mode...
2025.12
7.6
0
Original
Model Role=Target Mode...
2025.12
6.6
-
Feedback
Search any
task
Search any
task