995generated behavioral tests
44.1%official best score
0.4%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 | 0.4% | ok | 4/995 | $7.46 | 110 | 0.32h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 0.4% | ok | 4/995 | n/a | 76 | 0.27h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 0.4% | ok | 4/995 | n/a | 84 | 0.30h | 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
0.4% from 4/995 ProgramBench-scored tests. Raw public eval statuses: failure: 1003, passed: 329, skipped: 78.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
bc90bea37ab6tests.test_build.test_mangen_missing_output_dir (failure)bc90bea37ab6tests.test_compgen.test_compgen_kind_dir_prefix (failure)bc90bea37ab6tests.test_cli.test_export_missing_file_error (failure)bc90bea37ab6tests.test_edge_cases.test_build_duplicate_param_name_error (failure)bc90bea37ab6tests.test_compgen.test_argc_value_path_with_extensions (failure)bc90bea37ab6tests.test_build.test_build_simple_script_stdout (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 | Claude Sonnet 4.6 | Anthropic | 44.1% | $79.14 | 919 |
| 2 | GPT 5.5 (xhigh) | OpenAI | 42.8% | $13.46 | 111 |
| 3 | Claude Opus 4.6 | Anthropic | 39.8% | $30.14 | 435 |
| 4 | GPT 5.5 (high) | OpenAI | 31.5% | $5.08 | 46 |
| 5 | Claude Opus 4.7 (xhigh) | Anthropic | 28.0% | $6.90 | 121 |
| 6 | GPT 5.5 | OpenAI | 18.4% | $2.18 | 29 |
| 7 | Claude Haiku 4.5 | Anthropic | 10.4% | $1.47 | 175 |
| 8 | GPT 5.4 | OpenAI | 10.3% | $0.55 | 16 |
| 9 | Gemini 3 Flash | 8.8% | $0.21 | 63 | |
| 10 | Claude Opus 4.7 | Anthropic | 7.9% | $0.61 | 34 |
| 11 | GPT 5 mini | OpenAI | 7.3% | $0.02 | 20 |
| 12 | Gemini 3.1 Pro | 7.1% | $1.30 | 88 | |
| 13 | GPT 5.4 mini | OpenAI | 2.4% | $0.02 | 7 |