872generated behavioral tests
77.5%official best score
84.1%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 | 84.1% | ok | 733/872 | $7.34 | 89 | 0.31h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 41.1% | ok | 358/872 | n/a | 118 | 0.35h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 39.1% | ok | 341/872 | n/a | 111 | 0.34h | 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
84.1% from 733/872 ProgramBench-scored tests. Raw public eval statuses: failure: 141, passed: 903.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
18e5fb09a17atests.test_control_file_errors.test_control_file_escape_f_v (failure)18e5fb09a17atests.test_control_file_errors.test_control_file_char_range (failure)18e5fb09a17atests.test_encoding.test_utf8_copyright_symbols (failure)18e5fb09a17atests.test_errors.test_corrupted_font_file_invalid_magic (failure)18e5fb09a17atests.test_control_file_errors.test_control_file_invalid_g_command (failure)18e5fb09a17atests.test_errors.test_null_byte_in_input (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 | Claude Opus 4.6 | Anthropic | 77.5% | $18.00 | 279 |
| 2 | Claude Opus 4.7 (xhigh) | Anthropic | 76.7% | $82.10 | 502 |
| 3 | GPT 5.5 (xhigh) | OpenAI | 72.6% | $15.89 | 137 |
| 4 | Claude Opus 4.7 | Anthropic | 72.2% | $8.06 | 134 |
| 5 | GPT 5.5 (high) | OpenAI | 69.6% | $6.70 | 74 |
| 6 | Claude Sonnet 4.6 | Anthropic | 67.0% | $44.94 | 645 |
| 7 | GPT 5.5 | OpenAI | 45.5% | $2.94 | 37 |
| 8 | GPT 5.4 | OpenAI | 42.8% | $0.52 | 16 |
| 9 | Gemini 3.1 Pro | 40.5% | $1.73 | 90 | |
| 10 | Claude Haiku 4.5 | Anthropic | 37.4% | $1.43 | 196 |
| 11 | Gemini 3 Flash | 30.3% | $0.28 | 68 | |
| 12 | GPT 5.4 mini | OpenAI | 9.1% | $0.03 | 10 |
| 13 | GPT 5 mini | OpenAI | 2.3% | $0.03 | 17 |