SWE-MM

SWE-MM - multimodal software engineering benchmark. Evaluated on Claude Code harness using public dev split of SWE-bench Multimodal with modifications.

Category

Agentic Coding

Max Score

100.0

Score Type

percent

Active

Yes

Model Scores

Scores for SWE-MM

Model Score Date Verified
Qwen3.8-27B
100.00%
18.08.2026 Verified
Qwen3.6-27B
25.70
18.08.2026 Unverified
Qwen3.7-Plus
33.33%
18.08.2026 Unverified
Opus4.6 Max
10.85%
18.08.2026 Unverified