5319generated behavioral tests
73.8%official best score
21.5%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) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 21.5% | ok | 1143/5319 | n/a | 78 | 0.26h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 15.8% | ok | 840/5319 | $6.06 | 80 | 0.33h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 1.0% | ok | 52/5319 | n/a | 46 | 0.20h | 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
15.8% from 840/5319 ProgramBench-scored tests. Raw public eval statuses: failure: 5169, passed: 847, skipped: 1144.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
20bd8129dbc9eval.tests.test_proj.TestVerboseModes.test_very_verbose_V_ellipsoid_info (failure)20bd8129dbc9eval.tests.test_proj.TestCombinedFlags.test_format_and_echo (failure)20bd8129dbc9eval.tests.test_proj.TestErrorHandling.test_missing_projection_arg (failure)20bd8129dbc9eval.tests.test_proj.TestCoverageFocused.test_scale_S_with_inverse (failure)20bd8129dbc9eval.tests.test_proj.TestCoverageFocused.test_vprocess_error_on_bad_input (failure)20bd8129dbc9eval.tests.test_proj.TestBasicInvocation.test_no_args_prints_usage (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 5.4 | OpenAI | 73.8% | $0.47 | 13 |
| 2 | GPT 5.5 (high) | OpenAI | 29.8% | $3.79 | 33 |
| 3 | Claude Opus 4.7 | Anthropic | 24.2% | $1.45 | 35 |
| 4 | Gemini 3.1 Pro | 22.4% | $1.01 | 66 | |
| 5 | GPT 5.4 mini | OpenAI | 22.4% | $0.02 | 10 |
| 6 | GPT 5.5 (xhigh) | OpenAI | 22.3% | $9.69 | 86 |
| 7 | Claude Sonnet 4.6 | Anthropic | 22.3% | $19.26 | 427 |
| 8 | Gemini 3 Flash | 22.3% | $0.19 | 80 | |
| 9 | Claude Opus 4.6 | Anthropic | 11.1% | $14.08 | 288 |
| 10 | GPT 5.5 | OpenAI | 5.3% | $1.00 | 15 |
| 11 | Claude Opus 4.7 (xhigh) | Anthropic | 3.0% | $2.37 | 79 |
| 12 | Claude Haiku 4.5 | Anthropic | 1.4% | $1.14 | 173 |
| 13 | GPT 5 mini | OpenAI | 0.0% | $0.03 | 16 |