455generated behavioral tests
80.7%official best score
56.9%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) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 56.9% | ok | 259/455 | n/a | 75 | 0.28h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 52.5% | ok | 239/455 | n/a | 67 | 0.22h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 40.9% | ok | 186/455 | $3.64 | 82 | 0.21h | manifest · eval json · eval summary · usage audit |
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) · 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.
GPT 5.5 (xhigh) · 20260517T094700Z
40.9% from 186/455 ProgramBench-scored tests. Raw public eval statuses: failure: 29, not_run: 249, passed: 216, skipped: 1.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
453e9bab36fdeval.tests.test_basic_invocation.test_args_flag (failure)453e9bab36fdeval.tests.test_configuration.test_invalid_config_file (failure)453e9bab36fdeval.tests.test_utility.test_zsh_completion (failure)453e9bab36fdeval.tests.test_utility.test_powershell_completion (failure)453e9bab36fdeval.tests.test_utility.test_elvish_completion (failure)15eca2a4caf9tests.test_cli.test_args_flag_displays_parsed_arguments (not_run)
Official ProgramBench Results by Model
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | Claude Opus 4.6 | Anthropic | 80.7% | $19.07 | 374 |
| 2 | GPT 5.5 (high) | OpenAI | 76.5% | $3.81 | 48 |
| 3 | Claude Opus 4.7 (xhigh) | Anthropic | 65.1% | $10.97 | 186 |
| 4 | Claude Opus 4.7 | Anthropic | 64.0% | $2.87 | 81 |
| 5 | GPT 5.5 | OpenAI | 61.8% | $0.78 | 15 |
| 6 | Gemini 3.1 Pro | 59.3% | $2.01 | 143 | |
| 7 | Gemini 3 Flash | 54.5% | $0.25 | 95 | |
| 8 | GPT 5.4 | OpenAI | 54.3% | $0.50 | 13 |
| 9 | Claude Haiku 4.5 | Anthropic | 49.7% | $1.32 | 196 |
| 10 | GPT 5.5 (xhigh) | OpenAI | 44.0% | $10.63 | 108 |
| 11 | Claude Sonnet 4.6 | Anthropic | 37.4% | $21.46 | 460 |
| 12 | GPT 5.4 mini | OpenAI | 36.7% | $0.05 | 10 |
| 13 | GPT 5 mini | OpenAI | 30.5% | $0.07 | 22 |