692generated behavioral tests
94.2%official best score
80.1%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 | 80.1% | ok | 554/692 | $2.93 | 49 | 0.16h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 10.0% | ok | 69/692 | n/a | 82 | 0.19h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 10.0% | ok | 69/692 | n/a | 89 | 0.24h | 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
80.1% from 554/692 ProgramBench-scored tests. Raw public eval statuses: failure: 168, passed: 611.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
77b08cf4e301eval.tests.test_config_management.test_invalid_toml_config (failure)77b08cf4e301eval.tests.test_cli_combinations.test_run_command_variations (failure)77b08cf4e301eval.tests.test_coverage_improvements.test_invalid_toml_syntax_error (failure)77b08cf4e301eval.tests.test_coverage_improvements.test_script_with_multiline_command (failure)77b08cf4e301eval.tests.test_coverage_improvements.test_run_with_env_variables (failure)77b08cf4e301eval.tests.test_edge_cases.test_copy_preserves_all_metadata (failure)
GPT 5.5 (xhigh) · 20260518T232006Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260519T215046Z
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 | 94.2% | $8.76 | 101 |
| 2 | Claude Opus 4.7 (xhigh) | Anthropic | 92.9% | $9.65 | 182 |
| 3 | GPT 5.5 (high) | OpenAI | 90.0% | $4.17 | 59 |
| 4 | Claude Opus 4.6 | Anthropic | 83.7% | $9.16 | 245 |
| 5 | Claude Opus 4.7 | Anthropic | 83.4% | $3.71 | 83 |
| 6 | GPT 5.5 | OpenAI | 76.3% | $1.20 | 18 |
| 7 | Gemini 3.1 Pro | 69.8% | $1.10 | 73 | |
| 8 | GPT 5.4 | OpenAI | 57.5% | $0.52 | 13 |
| 9 | Gemini 3 Flash | 52.7% | $0.35 | 133 | |
| 10 | Claude Haiku 4.5 | Anthropic | 45.7% | $0.88 | 115 |
| 11 | GPT 5.4 mini | OpenAI | 36.8% | $0.08 | 16 |
| 12 | GPT 5 mini | OpenAI | 14.6% | $0.03 | 29 |
| 13 | Claude Sonnet 4.6 | Anthropic | 2.9% | $4.77 | 211 |