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 | 46.3% | ok | 1538/3322 | $6.81 | 67 | 0.27h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 45.5% | ok | 1510/3322 | $5.98 | 76 | 0.24h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 2.7% | warn 1 | 91/3323 | $4.96 | 56 | 0.25h | 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
45.5% from 1510/3322 ProgramBench-scored tests. Raw public eval statuses: failure: 2042, passed: 2601, skipped: 29.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
e5d3d967093eeval.tests.test_additional_rules.test_rule_with_custom_config (failure)e5d3d967093eeval.tests.test_additional_features.test_completions_fish (failure)e5d3d967093eeval.tests.test_additional_features.test_check_with_config_file (failure)e5d3d967093eeval.tests.test_additional_rules.test_md061_forbidden_terms (failure)e5d3d967093eeval.tests.test_check_command.test_check_statistics_flag (failure)e5d3d967093eeval.tests.test_check_command.test_check_nonexistent_file (failure)
GPT 5.5 (xhigh) · 20260518T232006Z
2.7% from 91/3323 ProgramBench-scored tests. Raw public eval statuses: error: 1, failure: 3584, not_run: 18, passed: 1036, skipped: 30.
Agent trace: 112 shell command(s), 72 executable-observation command(s), 5 build command(s), 2 blocked attempt(s). Raw Codex logs are not published.
Likely miss class: behavioral mismatch, exact version/output mismatch.
e5d3d967093eeval.tests.test_additional_rules.test_fix_with_specific_rules (failure)e5d3d967093eeval.tests.test_all_rules_comprehensive.test_multiple_files_with_different_violations (failure)e5d3d967093eeval.tests.test_additional_features.test_check_show_full_path (failure)e5d3d967093eeval.tests.test_advanced_rules.test_statistics_output (failure)e5d3d967093eeval.tests.test_all_rules_comprehensive.test_md010_hard_tabs (failure)e5d3d967093eeval.tests.test_advanced_rules.test_md047_single_trailing_newline (failure)
manifest · eval json · eval summary · agent summary · usage audit
Official ProgramBench Results by Model
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | GPT 5.5 (high) | OpenAI | 48.6% | $4.29 | 32 |
| 2 | GPT 5.5 (xhigh) | OpenAI | 48.2% | $9.73 | 63 |
| 3 | Claude Opus 4.6 | Anthropic | 40.7% | $8.20 | 206 |
| 4 | GPT 5.5 | OpenAI | 37.9% | $1.37 | 16 |
| 5 | Claude Opus 4.7 (xhigh) | Anthropic | 34.1% | $3.78 | 98 |
| 6 | Gemini 3 Flash | 25.9% | $0.34 | 77 | |
| 7 | Gemini 3.1 Pro | 23.5% | $1.22 | 66 | |
| 8 | Claude Opus 4.7 | Anthropic | 14.7% | $1.36 | 41 |
| 9 | GPT 5.4 | OpenAI | 10.8% | $0.36 | 7 |
| 10 | GPT 5 mini | OpenAI | 4.9% | $0.04 | 21 |
| 11 | Claude Sonnet 4.6 | Anthropic | 2.8% | $63.87 | 785 |
| 12 | Claude Haiku 4.5 | Anthropic | 1.4% | $0.60 | 88 |
| 13 | GPT 5.4 mini | OpenAI | 0.0% | $0.06 | 13 |