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 | 68.4% | ok | 403/589 | n/a | 41 | 0.13h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 67.2% | ok | 396/589 | n/a | 39 | 0.11h | manifest · eval json · eval summary · agent summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 65.4% | ok | 385/589 | $3.89 | 54 | 0.20h | manifest · eval json · eval 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) · 20260518T232006Z
67.2% from 396/589 ProgramBench-scored tests. Raw public eval statuses: error: 4, failure: 204, passed: 519, skipped: 35.
Agent trace: 59 shell command(s), 37 executable-observation command(s), 9 build command(s), 2 blocked attempt(s). Raw Codex logs are not published.
Likely miss class: behavioral mismatch, exact version/output mismatch.
db975e1644f7tests.test_completions.test_completions_case_sensitive_powershell_mixed (failure)db975e1644f7tests.test_cli_basics.test_completions_invalid_shell_error (failure)db975e1644f7tests.test_clean_errors.test_version_with_quiet_flag (failure)db975e1644f7tests.test_clean_errors.test_clean_multiple_times_in_row (failure)db975e1644f7tests.test_cli_basics.test_flags_order_independence_quiet_version (failure)db975e1644f7tests.test_clean_errors.test_clean_preserves_other_target_directories (failure)
manifest · eval json · eval summary · agent summary · usage audit
GPT 5.5 (xhigh) · 20260517T094700Z
65.4% from 385/589 ProgramBench-scored tests. Raw public eval statuses: error: 4, failure: 212, passed: 512, skipped: 34.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
db975e1644f7tests.test_completions.test_completions_case_sensitive_powershell_mixed (failure)db975e1644f7tests.test_completions.test_completions_powershell_golden_exact (failure)db975e1644f7tests.test_cli_basics.test_completions_bash_generates_valid_script (failure)db975e1644f7tests.test_completions.test_completions_bash_golden_exact (failure)db975e1644f7tests.test_cli_basics.test_no_arguments_lsp_mode_message (failure)db975e1644f7tests.test_clean_errors.test_completions_all_shells_generate_output (failure)
Official ProgramBench Results by Model
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | GPT 5.5 (high) | OpenAI | 88.5% | $2.86 | 38 |
| 2 | GPT 5.5 (xhigh) | OpenAI | 78.3% | $5.00 | 49 |
| 3 | Claude Sonnet 4.6 | Anthropic | 75.2% | $10.43 | 237 |
| 4 | Claude Opus 4.6 | Anthropic | 71.8% | $10.74 | 250 |
| 5 | GPT 5.5 | OpenAI | 66.9% | $0.67 | 14 |
| 6 | Claude Opus 4.7 (xhigh) | Anthropic | 65.7% | $4.86 | 122 |
| 7 | Claude Opus 4.7 | Anthropic | 62.8% | $2.22 | 84 |
| 8 | Gemini 3.1 Pro | 56.4% | $1.96 | 123 | |
| 9 | Claude Haiku 4.5 | Anthropic | 44.1% | $0.65 | 108 |
| 10 | GPT 5.4 | OpenAI | 42.4% | $0.25 | 10 |
| 11 | Gemini 3 Flash | 39.2% | $0.29 | 65 | |
| 12 | GPT 5 mini | OpenAI | 29.2% | $0.02 | 19 |
| 13 | GPT 5.4 mini | OpenAI | 25.3% | $0.03 | 7 |