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 | 17.1% | ok | 109/638 | $3.08 | 48 | 0.20h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 14.9% | ok | 95/638 | $2.50 | 44 | 0.15h | manifest · eval json · eval summary · agent summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 13.0% | ok | 83/638 | $5.66 | 69 | 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
17.1% from 109/638 ProgramBench-scored tests. Raw public eval statuses: error: 435, failure: 627, passed: 145, skipped: 16.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
803faac7d834eval.tests.test_coverage_targeted.TestCompressedState.test_compressed_state_saves_and_restores (failure)803faac7d834eval.tests.test_additional_coverage.TestEditFilesModeEdgeCases.test_edit_files_mode_empty_path_fails (failure)803faac7d834eval.tests.test_coverage_targeted.TestEditTomlMode.test_edit_in_toml_mode (failure)803faac7d834eval.tests.test_coverage_boost.TestLogJsonVariants.test_log_json_all (error)803faac7d834eval.tests.test_additional_coverage.TestOrderByStatusCoverage.test_order_by_status_with_mixed_statuses (error)803faac7d834eval.tests.test_coverage_targeted.TestCleanCommand.test_clean_successful_only_keeps_failed (error)
GPT 5.5 (xhigh) · 20260518T232006Z
14.9% from 95/638 ProgramBench-scored tests. Raw public eval statuses: error: 435, failure: 641, passed: 131, skipped: 16.
Agent trace: 76 shell command(s), 60 executable-observation command(s), 4 build command(s), 3 blocked attempt(s). Raw Codex logs are not published.
Likely miss class: behavioral mismatch.
803faac7d834eval.tests.test_additional_coverage.TestEditFilesModeEdgeCases.test_edit_files_mode_empty_path_fails (failure)803faac7d834eval.tests.test_coverage_targeted.TestCompressedState.test_compressed_state_saves_and_restores (failure)803faac7d834eval.tests.test_coverage_targeted.TestEditTomlMode.test_edit_in_toml_mode (failure)803faac7d834eval.tests.test_basic_invocation.TestCompletions.test_completions_to_nonexistent_dir (failure)803faac7d834eval.tests.test_additional_coverage.TestEditCoverage.test_edit_multiple_tasks (error)803faac7d834eval.tests.test_coverage_boost.TestQueryDescendingOrder.test_order_by_command_desc (error)
manifest · eval json · eval summary · agent summary · usage audit
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 Opus 4.7 (xhigh) | Anthropic | 16.1% | $7.48 | 127 |
| 2 | Claude Opus 4.6 | Anthropic | 15.4% | $12.90 | 252 |
| 3 | GPT 5.5 | OpenAI | 13.6% | $1.09 | 14 |
| 4 | Gemini 3.1 Pro | 12.5% | $1.07 | 72 | |
| 5 | Claude Opus 4.7 | Anthropic | 12.4% | $1.93 | 68 |
| 6 | GPT 5.5 (high) | OpenAI | 8.6% | $3.36 | 45 |
| 7 | GPT 5.5 (xhigh) | OpenAI | 8.3% | $8.76 | 87 |
| 8 | Claude Haiku 4.5 | Anthropic | 4.2% | $0.56 | 118 |
| 9 | GPT 5.4 mini | OpenAI | 4.1% | $0.04 | 9 |
| 10 | GPT 5.4 | OpenAI | 3.9% | $0.16 | 7 |
| 11 | Gemini 3 Flash | 2.8% | $0.30 | 91 | |
| 12 | GPT 5 mini | OpenAI | 1.9% | $0.02 | 11 |
| 13 | Claude Sonnet 4.6 | Anthropic | 1.6% | $15.46 | 231 |