874generated behavioral tests
93.0%official best score
85.4%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 | 85.4% | ok | 746/874 | $4.01 | 59 | 0.25h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 85.1% | ok | 744/874 | n/a | 50 | 0.17h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 81.5% | ok | 712/874 | n/a | 53 | 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
85.4% from 746/874 ProgramBench-scored tests. Raw public eval statuses: failure: 306, passed: 914, skipped: 24.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
63ece4f28ebdtests.test_50_percent.TestMoreLsCombinations.test_ls_bytes_apparent_count (failure)63ece4f28ebdtests.test_absolutely_50.TestFinal.test_ls_full_path_all_dirs (failure)63ece4f28ebdtests.test_additional_coverage_phase2.TestCGIOptions.test_cgi_with_footer (failure)63ece4f28ebdtests.test_50_percent.TestHistogramAllCombinations.test_histogram_all_flags (failure)63ece4f28ebdtests.test_command_execution.TestDatabaseEnvironmentVariable.test_index_uses_env_database (failure)63ece4f28ebdtests.test_additional_coverage_phase2.TestManualSubcommand.test_manual_command_exists (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 | 93.0% | $9.70 | 86 |
| 2 | Claude Opus 4.7 (xhigh) | Anthropic | 87.6% | $21.07 | 218 |
| 3 | Claude Opus 4.7 | Anthropic | 83.4% | $2.14 | 201 |
| 4 | GPT 5.5 (high) | OpenAI | 80.0% | $3.81 | 40 |
| 5 | GPT 5.5 | OpenAI | 71.1% | $1.19 | 17 |
| 6 | GPT 5.4 | OpenAI | 40.7% | $0.30 | 7 |
| 7 | GPT 5.4 mini | OpenAI | 25.1% | $0.06 | 7 |
| 8 | Gemini 3 Flash | 23.9% | $0.40 | 78 | |
| 9 | GPT 5 mini | OpenAI | 16.8% | $0.02 | 13 |
| 10 | Claude Haiku 4.5 | Anthropic | 16.7% | $1.33 | 155 |
| 11 | Claude Sonnet 4.6 | Anthropic | 11.8% | $30.03 | 441 |
| 12 | Claude Opus 4.6 | Anthropic | 0.0% | $5.00 | 228 |
| 13 | Gemini 3.1 Pro | 0.0% | $2.24 | 142 |