229generated behavioral tests
34.5%official best score
34.1%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 | 34.1% | ok | 78/229 | $3.11 | 50 | 0.15h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 31.9% | ok | 73/229 | n/a | 108 | 0.27h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 1.3% | warn 1 | 3/230 | n/a | 61 | 0.17h | 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
34.1% from 78/229 ProgramBench-scored tests. Raw public eval statuses: failure: 157, passed: 93, skipped: 1.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
02019de415b5eval.tests.test_argparse_validation.test_debug_flag_requires_known_specifier (failure)02019de415b5eval.tests.test_argparse_validation.test_missing_or_invalid_format_specifiers[args3-expected3] (failure)02019de415b5eval.tests.test_argparse_validation.test_missing_or_invalid_format_specifiers[args1-expected1] (failure)8e06264b11b3tests.test_mode_dispatch.TestErrorHandling.test_feature_mode_missing_specifier (failure)8e06264b11b3tests.test_mode_dispatch.TestDebugModes.test_debug_level_recognized (failure)ffcc3ee94dc7eval.tests.test_doxygen_testing_suite_externalized.test_ext_internal_dox_test_via_cli[001_001_a.dox] (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 | Claude Opus 4.7 | Anthropic | 34.5% | $1.17 | 50 |
| 2 | GPT 5.4 | OpenAI | 29.7% | $0.24 | 7 |
| 3 | Claude Opus 4.7 (xhigh) | Anthropic | 21.7% | $5.38 | 138 |
| 4 | GPT 5.5 | OpenAI | 21.0% | $1.45 | 20 |
| 5 | GPT 5.5 (high) | OpenAI | 20.5% | $2.41 | 29 |
| 6 | Claude Haiku 4.5 | Anthropic | 17.8% | $0.73 | 130 |
| 7 | GPT 5 mini | OpenAI | 16.6% | $0.02 | 11 |
| 8 | Gemini 3.1 Pro | 15.7% | $1.59 | 87 | |
| 9 | GPT 5.4 mini | OpenAI | 14.8% | $0.02 | 6 |
| 10 | GPT 5.5 (xhigh) | OpenAI | 1.7% | $6.04 | 51 |
| 11 | Claude Opus 4.6 | Anthropic | 1.3% | $15.13 | 328 |
| 12 | Claude Sonnet 4.6 | Anthropic | 1.3% | $28.02 | 525 |
| 13 | Gemini 3 Flash | 1.3% | $0.21 | 76 |