Codex Results by Model
| # | Model | Run | Agent | Mode | Score | Eval | Tests | Est. cost | Calls | Wall | Evidence |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 92.0% | ok | 496/539 | $3.31 | 47 | 0.21h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 62.3% | ok | 336/539 | $2.17 | 39 | 0.14h | manifest · eval json · eval summary · agent summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 61.6% | ok | 332/539 | $3.63 | 55 | 0.22h | not exported |
Why Scores Differ
Official ProgramBench rows are the public mini-SWE-agent submissions. GoalBench rows are separate Codex /goal submissions. Same model label does not mean the same agent, prompt, tool loop, or generated implementation. A GoalBench row is resolved only when the ProgramBench-scored pass rate is exactly 100%.
GPT 5.5 (xhigh) · 20260517T094700Z
92.0% from 496/539 ProgramBench-scored tests. Raw public eval statuses: failure: 112, passed: 638.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
d161b26b57a6eval.tests.test_i3_style.test_to_theme_with_config_flag (failure)1ae1a446d287tests.test_all_builtin_themes.test_builtin_theme_application[oceanic-next] (failure)1ae1a446d287tests.test_config_to_theme.test_to_theme_with_config_flag (failure)1ae1a446d287tests.test_all_builtin_themes.test_builtin_theme_application[ubuntu] (failure)1ae1a446d287tests.test_all_builtin_themes.test_builtin_theme_application[purple] (failure)1ae1a446d287tests.test_all_builtin_themes.test_builtin_theme_application[tomorrow-night-80s] (failure)
GPT 5.5 (xhigh) · 20260518T232006Z
62.3% from 336/539 ProgramBench-scored tests. Raw public eval statuses: failure: 282, passed: 468.
Agent trace: 64 shell command(s), 43 executable-observation command(s), 2 build command(s), 0 blocked attempt(s). Raw Codex logs are not published.
Likely miss class: behavioral mismatch, exact version/output mismatch.
d161b26b57a6eval.tests.test_i3_style.test_custom_theme_from_file (failure)d161b26b57a6eval.tests.test_i3_style.test_custom_config_with_c_flag (failure)d161b26b57a6eval.tests.test_i3_style.test_custom_config_with_config_flag (failure)d161b26b57a6eval.tests.test_i3_style.test_nonexistent_config_file (failure)d161b26b57a6eval.tests.test_i3_style.test_stdout_output_default (failure)d161b26b57a6eval.tests.test_i3_style.test_output_to_file_with_o_flag (failure)
manifest · eval json · eval summary · agent summary · usage audit
GPT 5.5 (xhigh) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
Official ProgramBench Results by Model
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | GPT 5.5 (xhigh) | OpenAI | 93.3% | $6.85 | 67 |
| 2 | Claude Opus 4.7 (xhigh) | Anthropic | 92.2% | $18.42 | 206 |
| 3 | GPT 5.5 | OpenAI | 87.2% | $2.21 | 28 |
| 4 | GPT 5.5 (high) | OpenAI | 87.0% | $4.41 | 43 |
| 5 | Claude Opus 4.6 | Anthropic | 80.0% | $12.87 | 205 |
| 6 | Claude Sonnet 4.6 | Anthropic | 77.9% | $16.07 | 322 |
| 7 | Claude Opus 4.7 | Anthropic | 73.5% | $7.30 | 165 |
| 8 | GPT 5.4 | OpenAI | 48.8% | $0.31 | 8 |
| 9 | Gemini 3 Flash | 48.8% | $0.53 | 107 | |
| 10 | Gemini 3.1 Pro | 42.7% | $2.02 | 133 | |
| 11 | GPT 5.4 mini | OpenAI | 35.4% | $0.02 | 8 |
| 12 | Claude Haiku 4.5 | Anthropic | 34.9% | $0.85 | 117 |
| 13 | GPT 5 mini | OpenAI | 30.2% | $0.02 | 11 |