1312generated behavioral tests
97.2%official best score
88.5%Codex best score
3Codex result rows
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 | 88.5% | ok | 1161/1312 | $7.01 | 115 | 0.35h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 86.4% | ok | 1133/1312 | n/a | 58 | 0.25h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 60.7% | ok | 796/1312 | n/a | 77 | 0.27h | 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
88.5% from 1161/1312 ProgramBench-scored tests. Raw public eval statuses: failure: 151, passed: 1367.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
0cc73f2d7c79eval.tests.test_complex_scenarios.test_surrogates_verbose_colorize (failure)0cc73f2d7c79eval.tests.test_coverage_boost.TestComplexFlagCombinations.test_escape_verbose_colorize_repetitions (failure)0cc73f2d7c79eval.tests.test_escaping.test_with_surrogates_without_escape_fails (failure)0cc73f2d7c79eval.tests.test_dependency_coverage.TestFileInputStress.test_file_with_long_lines (failure)0cc73f2d7c79eval.tests.test_extreme_coverage.TestMegaCombinations.test_mega_combo_2 (failure)0cc73f2d7c79eval.tests.test_extreme_coverage.TestColorizeVariations.test_colorize_escape_repetitions (failure)
GPT 5.5 (xhigh) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260518T232006Z
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 | 97.2% | $9.20 | 71 |
| 2 | GPT 5.5 (high) | OpenAI | 88.7% | $6.81 | 57 |
| 3 | Claude Opus 4.7 (xhigh) | Anthropic | 82.1% | $31.59 | 231 |
| 4 | Claude Opus 4.7 | Anthropic | 73.9% | $7.62 | 153 |
| 5 | GPT 5.4 | OpenAI | 66.5% | $0.32 | 7 |
| 6 | Claude Opus 4.6 | Anthropic | 66.0% | $16.93 | 266 |
| 7 | Gemini 3.1 Pro | 61.4% | $1.29 | 159 | |
| 8 | GPT 5.5 | OpenAI | 58.1% | $1.56 | 22 |
| 9 | Claude Sonnet 4.6 | Anthropic | 56.3% | $48.82 | 743 |
| 10 | Gemini 3 Flash | 52.8% | $0.23 | 81 | |
| 11 | GPT 5 mini | OpenAI | 34.7% | $0.08 | 36 |
| 12 | GPT 5.4 mini | OpenAI | 21.6% | $0.05 | 10 |
| 13 | Claude Haiku 4.5 | Anthropic | 0.0% | $1.02 | 104 |