GSO evaluates SWE agents on 102 real software optimization tasks spanning 10 codebases and five programming languages. OPT@1 is the share of tasks where one attempt passes correctness tests and reaches at least 95% of the human expert speedup.
Browse the latest scores, model modes, release dates, and parameter sizes for GSO.
Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology
| Rank | Model | License | |||
|---|---|---|---|---|---|
 Opus 4.7 Standard ModeTools | 44.12 | 2026-04-16 | Unknown | Closed | |
Opus 4.7 Thinking Level · HighTools | 44.10 | 2026-04-16 | Unknown | Closed | |
Claude Opus 4.6 Thinking Level · HighTools | 41.20 | 2026-02-05 | Unknown | Closed | |
4 |  GPT-5.5 Thinking Level · Extra HighTools | 40.20 | 2026-04-23 | Unknown | Closed |
5 | Claude Opus 4.6 Standard ModeTools | 33.33 | 2026-02-05 | Unknown | Closed |
6 | GPT-5.4 Thinking Level · Extra HighTools | 31.37 | 2026-03-05 | Unknown | Closed |
7 | GPT-5.2 Thinking Level · HighTools | 27.40 | 2025-12-11 | Unknown | Closed |
8 | Opus 4.5 Standard ModeTools | 26.50 | 2025-11-25 | Unknown | Closed |
9 | GPT-5.4 Thinking Level · HighTools | 25.49 | 2026-03-05 | Unknown | Closed |
10 |  Gemini 3.1 Pro Preview Standard ModeTools | 22.55 | 2026-02-20 | Unknown | Closed |
11 | Gemini 3 Pro Standard ModeTools | 18.60 | 2026-02-01 | Unknown | Closed |
12 | Claude Sonnet 4.5 Standard ModeTools | 14.70 | 2025-09-30 | Unknown | Closed |
13 | GPT-5.1 Thinking Level · HighTools | 13.70 | 2025-11-12 | Unknown | Closed |
14 | Gemini 3.0 Flash Standard ModeTools | 9.80 | 2025-12-17 | Unknown | Closed |
15 | OpenAI o3 Thinking Level · HighTools | 8.80 | 2025-04-16 | Unknown | Closed |
16 | GPT-5 Thinking Level · HighTools | 6.90 | 2025-08-07 | Unknown | Closed |
17 | Claude 3.5 Sonnet New Standard ModeTools | 4.60 | 2024-10-22 | Unknown | Closed |
18 | Gemini 2.5-Pro Standard ModeTools | 3.92 | 2025-06-05 | Unknown | Closed |
19 | Claude Sonnet 3.7 Standard ModeTools | 3.80 | 2025-02-25 | Unknown | Closed |
20 | OpenAI o4 - mini Thinking Level · HighTools | 3.60 | 2025-04-16 | Unknown | Closed |
21 | GPT-4o(2024-11-20) Standard ModeTools | 0.00 | 2024-11-20 | Unknown | Closed |
In the official leaderboard snapshot observed on 2026-08-12, Claude Opus 4.8 with OpenHands reached 47.06% OPT@1.
