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 | 57.3% | ok | 150/262 | $2.59 | 49 | 0.13h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 50.8% | ok | 133/262 | $3.02 | 59 | 0.15h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 49.2% | ok | 129/262 | $1.91 | 37 | 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) · 20260517T094700Z
57.3% from 150/262 ProgramBench-scored tests. Raw public eval statuses: failure: 175, passed: 234, skipped: 4.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, interactive terminal behavior mismatch, terminal rendering mismatch.
222dd9bb6f48eval.tests.test_basic.test_no_arguments_requires_git_repo (failure)222dd9bb6f48eval.tests.test_tui_startup.test_tui_starts_and_quits (failure)222dd9bb6f48eval.tests.test_tui_startup.test_tui_help_toggle (failure)222dd9bb6f48eval.tests.test_tui_startup.test_tui_with_branches (failure)222dd9bb6f48eval.tests.test_tui_startup.test_tui_refresh_command (failure)222dd9bb6f48eval.tests.test_tui_startup.test_tui_navigation_keys (failure)
GPT 5.5 (xhigh) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260518T232006Z
49.2% from 129/262 ProgramBench-scored tests. Raw public eval statuses: failure: 192, passed: 214, skipped: 4.
Agent trace: 75 shell command(s), 49 executable-observation command(s), 4 build command(s), 2 blocked attempt(s). Raw Codex logs are not published.
Likely miss class: behavioral mismatch, interactive terminal behavior mismatch, terminal rendering mismatch.
222dd9bb6f48eval.tests.test_basic.test_no_arguments_requires_git_repo (failure)222dd9bb6f48eval.tests.test_tui_startup.test_tui_starts_and_quits (failure)222dd9bb6f48eval.tests.test_tui_startup.test_tui_help_toggle (failure)222dd9bb6f48eval.tests.test_tui_startup.test_tui_with_branches (failure)222dd9bb6f48eval.tests.test_tui_startup.test_tui_refresh_command (failure)222dd9bb6f48eval.tests.test_tui_startup.test_tui_with_merge_commit (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 | 87.0% | $7.34 | 70 |
| 2 | GPT 5.5 (high) | OpenAI | 75.2% | $3.94 | 50 |
| 3 | Claude Opus 4.7 | Anthropic | 67.2% | $1.11 | 48 |
| 4 | GPT 5.5 | OpenAI | 66.4% | $1.02 | 18 |
| 5 | Claude Opus 4.7 (xhigh) | Anthropic | 65.3% | $3.33 | 79 |
| 6 | GPT 5.4 | OpenAI | 36.6% | $0.14 | 9 |
| 7 | Gemini 3 Flash | 26.0% | $0.28 | 64 | |
| 8 | GPT 5.4 mini | OpenAI | 21.8% | $0.10 | 142 |
| 9 | GPT 5 mini | OpenAI | 16.0% | $0.01 | 9 |
| 10 | Claude Haiku 4.5 | Anthropic | 10.3% | $0.64 | 109 |
| 11 | Gemini 3.1 Pro | 7.3% | $2.32 | 92 | |
| 12 | Claude Opus 4.6 | Anthropic | 0.0% | $2.19 | 178 |
| 13 | Claude Sonnet 4.6 | Anthropic | 0.0% | $5.25 | 270 |