AgentWorldBench SWE

AgentWorldBench SWE domain - open-ended rubric evaluation for language world models simulating software engineering environments. Normalized to 0-100.

Category

General Agents

Max Score

100.0

Score Type

percent

Active

Yes

Model Scores

Scores for AgentWorldBench SWE

Model Score Date Verified
Qwen-AgentWorld-35B-A3B
90.79%
24.06.2026 Verified
Qwen-AgentWorld-397B-A17B
100.00%
24.06.2026 Unverified
Qwen3.5-35B-A3B
32.66%
24.06.2026 Unverified
Qwen3.5-397B-A17B
86.96%
24.06.2026 Unverified
Qwen3.6-Plus
69.69%
24.06.2026 Unverified
GPT-5.4
92.91%
24.06.2026 Unverified
Claude Opus 4.8
85.86%
24.06.2026 Unverified
Claude Opus 4.6
87.31%
24.06.2026 Unverified
Gemini 3.1 Pro
69.66%
24.06.2026 Unverified
Claude Sonnet 4.6
87.21%
24.06.2026 Unverified
DeepSeek-V4-Pro
70.85%
24.06.2026 Unverified
GLM-5.1
47.12%
24.06.2026 Unverified
Kimi K2.6
68.70%
24.06.2026 Unverified
MiniMax-M2.7
37.44
24.06.2026 Unverified