SWE Atlas - TW

SWE Atlas TW - agentic coding subtask; mini SWE agent harness, temp=1.0, top_p=0.95, 128K context, avg of 5 runs.

Category

Agentic Coding

Max Score

100.0

Score Type

percent

Active

Yes

Model Scores

Scores for SWE Atlas - TW

Model Score Date Verified
GPT 5.5
59.41%
11.06.2026 Unverified
MiniMax-M2.7
22.55%
11.06.2026 Unverified
Ornith-1.0-9B
16.95%
25.06.2026 Verified
Qwen3.5-9B
4.40
25.06.2026 Unverified
Qwen3.5-35B-A3B
8.40%
25.06.2026 Unverified
Ornith-1.0-35B
36.39%
25.06.2026 Verified
Qwen3.6-35B-A3B
13.84%
25.06.2026 Unverified
Qwen3.5-397B-A17B
21.93%
25.06.2026 Unverified
Claude Sonnet 4.6
42.61%
11.06.2026 Unverified
Claude Opus 4.7
52.57%
11.06.2026 Unverified
MiniMax-M3
41.06%
11.06.2026 Unverified
Gemini 3.1 Pro
39.50%
11.06.2026 Unverified
Hy3
48.99%
28.08.2026 Verified
Hy4 preview
83.05%
28.08.2026 Verified
DeepSeek V4 Pro
64.07%
28.08.2026 Verified
Qwen3.8-2.4T-A95B
75.27%
28.08.2026 Verified
GLM-5.3
70.30%
28.08.2026 Verified
Kimi K3
48.52%
28.08.2026 Verified
GPT-5.6 Sol
70.30%
28.08.2026 Verified
Claude Opus 5
100.00%
28.08.2026 Verified