3115generated behavioral tests
76.4%official best score
68.3%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) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 68.3% | ok | 2129/3115 | n/a | 70 | 0.27h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 67.5% | ok | 2103/3115 | $6.60 | 95 | 0.29h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 19.4% | ok | 603/3115 | n/a | 68 | 0.25h | 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) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260517T094700Z
67.5% from 2103/3115 ProgramBench-scored tests. Raw public eval statuses: failure: 1042, passed: 2325, skipped: 68.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
825a526c70aeeval.tests.test_config_advanced.TestCommandLineFlags.test_no_error_flag_suppresses_errors (failure)825a526c70aeeval.tests.test_basic.TestCompletionHelpers.test_root_options_completion (failure)825a526c70aeeval.tests.test_coverage_boost.TestSpecialSearchPatterns.test_like_pattern_with_underscore (failure)825a526c70aeeval.tests.test_acl_functions.TestAclColumn.test_acl_column_shows_acl_entries (failure)825a526c70aeeval.tests.test_coverage_boost2.TestHgIgnoreEdgeCases.test_hgignore_with_syntax_directive (failure)825a526c70aeeval.tests.test_coverage_final.TestFunctionEdgeCases.test_format_time_non_numeric (failure)
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 | 76.4% | $11.81 | 82 |
| 2 | GPT 5.5 (high) | OpenAI | 64.9% | $3.56 | 39 |
| 3 | GPT 5.5 | OpenAI | 57.3% | $1.57 | 18 |
| 4 | Claude Opus 4.6 | Anthropic | 44.0% | $9.78 | 262 |
| 5 | Claude Opus 4.7 | Anthropic | 29.3% | $2.52 | 74 |
| 6 | Gemini 3.1 Pro | 24.1% | $0.89 | 62 | |
| 7 | GPT 5.4 mini | OpenAI | 20.3% | $0.05 | 9 |
| 8 | GPT 5.4 | OpenAI | 20.3% | $0.43 | 12 |
| 9 | Claude Haiku 4.5 | Anthropic | 9.9% | $0.51 | 113 |
| 10 | Claude Sonnet 4.6 | Anthropic | 2.5% | $20.11 | 479 |
| 11 | Claude Opus 4.7 (xhigh) | Anthropic | 1.4% | $7.60 | 126 |
| 12 | Gemini 3 Flash | 0.9% | $0.24 | 89 | |
| 13 | GPT 5 mini | OpenAI | 0.0% | $0.07 | 28 |