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 | 58.6% | ok | 468/799 | $3.99 | 54 | 0.20h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 46.1% | ok | 368/799 | $3.79 | 61 | 0.16h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 45.1% | ok | 360/799 | $4.62 | 71 | 0.24h | manifest · eval json · eval summary · agent 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) · 20260517T094700Z
58.6% from 468/799 ProgramBench-scored tests. Raw public eval statuses: failure: 336, passed: 604, skipped: 39.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
0fcd4ac3ed40eval.tests.test_bat.test_tabs_option (failure)0fcd4ac3ed40eval.tests.test_bat.test_bat_opts_env (failure)0fcd4ac3ed40eval.tests.test_bat.test_strip_ansi_always (failure)0fcd4ac3ed40eval.tests.test_bat.test_multiple_file_names (failure)0fcd4ac3ed40eval.tests.test_bat.test_invalid_line_range_format (failure)0fcd4ac3ed40eval.tests.test_bat.test_list_themes_colored (failure)
GPT 5.5 (xhigh) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260518T232006Z
45.1% from 360/799 ProgramBench-scored tests. Raw public eval statuses: failure: 447, passed: 493, skipped: 39.
Agent trace: 213 shell command(s), 182 executable-observation command(s), 6 build command(s), 1 blocked attempt(s). Raw Codex logs are not published.
Likely miss class: behavioral mismatch, exact version/output mismatch.
0fcd4ac3ed40eval.tests.test_bat.test_concatenate_file_stdin_file (failure)0fcd4ac3ed40eval.tests.test_bat.test_no_args_reads_stdin (failure)0fcd4ac3ed40eval.tests.test_bat.test_multiple_line_ranges (failure)0fcd4ac3ed40eval.tests.test_bat.test_style_header_filesize (failure)0fcd4ac3ed40eval.tests.test_bat.test_file_name_option (failure)0fcd4ac3ed40eval.tests.test_bat.test_quiet_empty_nonempty (failure)
manifest · eval json · eval summary · agent summary · usage audit
Official ProgramBench Results by Model
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | GPT 5.5 (xhigh) | OpenAI | 61.5% | $7.87 | 82 |
| 2 | Claude Sonnet 4.6 | Anthropic | 60.5% | $21.25 | 514 |
| 3 | GPT 5.5 (high) | OpenAI | 49.6% | $5.73 | 54 |
| 4 | Claude Opus 4.7 (xhigh) | Anthropic | 43.9% | $4.75 | 110 |
| 5 | GPT 5.5 | OpenAI | 40.6% | $1.25 | 18 |
| 6 | Gemini 3.1 Pro | 33.2% | $1.72 | 109 | |
| 7 | Gemini 3 Flash | 33.2% | $0.24 | 83 | |
| 8 | Claude Opus 4.7 | Anthropic | 31.9% | $1.16 | 48 |
| 9 | Claude Haiku 4.5 | Anthropic | 28.0% | $0.62 | 129 |
| 10 | GPT 5 mini | OpenAI | 28.0% | $0.02 | 11 |
| 11 | GPT 5.4 | OpenAI | 13.4% | $0.26 | 9 |
| 12 | GPT 5.4 mini | OpenAI | 3.1% | $0.03 | 8 |
| 13 | Claude Opus 4.6 | Anthropic | 0.0% | $2.85 | 253 |