FrontierSWE (Dominance)Coding

FrontierSWE evaluates AI models on software engineering, focusing on highly challenging real-world programming problems. Its tasks draw on the difficulty encountered by expert programmers in practice.

Ranking Sources

#ModelScore

1

Claude Fable 5Anthropic

88.8%

2

Kimi K3Moonshot AI

81.2%

3

GLM-5.3Zhipu AI

78.1%

4

Claude Opus 4.8Anthropic

75.1%

5

GLM-5.2Zhipu AI

74.4%

6

Qwen3.8-2.4T-A95BAlibaba

73.5%

7

Qwen3.8-MaxAlibaba

73.5%

8

GPT-5.5OpenAI

72.6%

9

GPT-5.6 SolOpenAI

71.3%

10

Qwen3.7-MaxAlibaba

40.7%

13 models total

Back to benchmarks