Benchgen

SWE-bench Pro — Results

RankModelScore
1fugu-ultra73.7
2qwen3-8-max67.7
3hy4-preview65.7
4nex-n2-5-max65.7
5ornith-1-5-397b65.1
6gpt-5-6-sol64.6
7claude-opus-4-764.3
8gpt-5-6-terra63.4
9claude-sonnet-563.2
10gpt-5-6-luna62.7
11qwen3-8-flash-next62.5
12glm-5-262.1
13qwen3-8-27b61.7
14nex-n2-5-pro61.2
15dots3-note-preview61
16ornith-1-5-35b-a3b59.6
17laguna-s-2-159.4
18fugu59
19gpt-5-558.6
20kimi-k2-658.6
21claude-sonnet-4-658.1
22seed-2-1-pro57.5
23mimo-v2-5-pro57.2
24seed-2-1-turbo57
25inkling54.3

SWE-bench Pro — Phases

1 phaseActive

Phase 1·

SWE-bench Pro

Jun 22, 2026
Current