810generated behavioral tests
97.4%official best score
91.5%Codex best score
3Codex result rows
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 | 91.5% | ok | 741/810 | $3.26 | 62 | 0.20h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 87.2% | ok | 706/810 | n/a | 47 | 0.17h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 28.3% | ok | 229/810 | n/a | 56 | 0.22h | 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
91.5% from 741/810 ProgramBench-scored tests. Raw public eval statuses: failure: 70, passed: 794, skipped: 5.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
55ccb3ccb510eval.tests.test_edge_cases.test_escape_tab (failure)55ccb3ccb510eval.tests.test_edge_cases.test_escape_backslash (failure)55ccb3ccb510eval.tests.test_edge_cases.test_escape_double_quote (failure)55ccb3ccb510eval.tests.test_edge_cases.test_escape_trailing_backslash (failure)55ccb3ccb510eval.tests.test_edge_cases.test_escape_unicode (failure)55ccb3ccb510eval.tests.test_preview.test_preview_multiple_files (failure)
GPT 5.5 (xhigh) · 20260518T232006Z
Public eval evidence has not been exported for this row yet.
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 | GPT 5.5 (xhigh) | OpenAI | 97.4% | $7.94 | 89 |
| 2 | Claude Opus 4.7 (xhigh) | Anthropic | 95.2% | $13.81 | 211 |
| 3 | Claude Opus 4.7 | Anthropic | 90.9% | $3.64 | 113 |
| 4 | GPT 5.5 (high) | OpenAI | 89.9% | $3.30 | 37 |
| 5 | Claude Opus 4.6 | Anthropic | 89.9% | $7.07 | 270 |
| 6 | GPT 5.5 | OpenAI | 86.0% | $0.87 | 17 |
| 7 | Claude Sonnet 4.6 | Anthropic | 85.1% | $3.76 | 181 |
| 8 | GPT 5.4 | OpenAI | 81.6% | $0.22 | 8 |
| 9 | Gemini 3 Flash | 80.7% | $0.25 | 111 | |
| 10 | Gemini 3.1 Pro | 80.0% | $1.33 | 122 | |
| 11 | GPT 5 mini | OpenAI | 70.9% | $0.03 | 22 |
| 12 | Claude Haiku 4.5 | Anthropic | 68.6% | $0.87 | 149 |
| 13 | GPT 5.4 mini | OpenAI | 45.8% | $0.04 | 14 |