Codex Results by Model
| # | Model | Run | Agent | Mode | Score | Eval | Tests | Est. cost | Calls | Wall | Evidence |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 85.5% | ok | 235/275 | $3.17 | 69 | 0.19h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 68.0% | ok | 187/275 | $4.31 | 75 | 0.22h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 65.8% | ok | 181/275 | $2.64 | 45 | 0.14h | 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) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260517T094700Z
68.0% from 187/275 ProgramBench-scored tests. Raw public eval statuses: error: 25, failure: 137, passed: 225, skipped: 4.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
b377895d0633eval.tests.test_basic_invocation.test_version_flag (failure)b377895d0633eval.tests.test_basic_invocation.test_version_format (failure)b377895d0633eval.tests.test_additional_coverage.test_bot_with_different_names (failure)b377895d0633eval.tests.test_bot_launch.test_bot_names_in_results (failure)b377895d0633eval.tests.test_basic_invocation.test_no_arguments_shows_error (failure)b377895d0633eval.tests.test_additional_coverage.test_multiple_flags_combined (failure)
GPT 5.5 (xhigh) · 20260518T232006Z
65.8% from 181/275 ProgramBench-scored tests. Raw public eval statuses: error: 25, failure: 139, passed: 223, skipped: 4.
Agent trace: 92 shell command(s), 52 executable-observation command(s), 5 build command(s), 2 blocked attempt(s). Raw Codex logs are not published.
Likely miss class: behavioral mismatch.
b377895d0633eval.tests.test_bot_launch.test_single_bot_game (failure)b377895d0633eval.tests.test_additional_coverage.test_very_small_map (failure)b377895d0633eval.tests.test_additional_coverage.test_multiple_flags_combined (failure)b377895d0633eval.tests.test_additional_coverage.test_seed_value_max_int (failure)b377895d0633eval.tests.test_additional_coverage.test_quiet_output_structure_complete (failure)b377895d0633eval.tests.test_bot_launch.test_bot_command_with_spaces (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 | 95.6% | $7.37 | 59 |
| 2 | GPT 5.5 (high) | OpenAI | 92.7% | $2.90 | 40 |
| 3 | Claude Opus 4.6 | Anthropic | 80.4% | $15.08 | 222 |
| 4 | GPT 5.5 | OpenAI | 80.0% | $1.06 | 17 |
| 5 | Gemini 3.1 Pro | 75.6% | $1.56 | 73 | |
| 6 | Claude Opus 4.7 | Anthropic | 71.3% | $1.00 | 44 |
| 7 | Gemini 3 Flash | 47.3% | $0.52 | 71 | |
| 8 | Claude Haiku 4.5 | Anthropic | 42.9% | $0.45 | 106 |
| 9 | Claude Opus 4.7 (xhigh) | Anthropic | 18.9% | $3.13 | 72 |
| 10 | GPT 5.4 | OpenAI | 18.5% | $0.20 | 6 |
| 11 | Claude Sonnet 4.6 | Anthropic | 14.2% | $38.69 | 494 |
| 12 | GPT 5 mini | OpenAI | 1.5% | $0.04 | 22 |
| 13 | GPT 5.4 mini | OpenAI | 0.0% | $0.02 | 9 |