1708generated behavioral tests
79.7%official best score
78.3%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 | 78.3% | ok | 1338/1708 | $7.71 | 104 | 0.30h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 66.3% | ok | 1133/1708 | n/a | 54 | 0.22h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 64.8% | ok | 1107/1708 | n/a | 77 | 0.19h | 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
78.3% from 1338/1708 ProgramBench-scored tests. Raw public eval statuses: error: 22, failure: 721, passed: 2262, skipped: 451.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
41d65330ce2ftests.test_command_line_parsing.test_password_prompt_flag_without_value (failure)41d65330ce2ftests.test_command_line_parsing.test_dash_dash_separator (failure)41d65330ce2ftests.test_error_conditions.test_init_with_system_schema (failure)41d65330ce2ftests.test_diff_error_paths.test_diff_with_malformed_config (failure)41d65330ce2ftests.test_format_error_paths.test_format_with_malformed_skeema_file (failure)41d65330ce2ftests.test_init_error_handling.test_init_with_invalid_environment_brackets (failure)
GPT 5.5 (xhigh) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260518T232006Z
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 | 79.7% | $9.34 | 94 |
| 2 | GPT 5.5 (high) | OpenAI | 78.5% | $4.82 | 55 |
| 3 | Claude Sonnet 4.6 | Anthropic | 76.5% | $16.73 | 416 |
| 4 | Claude Opus 4.6 | Anthropic | 74.9% | $7.91 | 269 |
| 5 | GPT 5.5 | OpenAI | 61.2% | $1.42 | 21 |
| 6 | Claude Opus 4.7 | Anthropic | 57.3% | $1.47 | 63 |
| 7 | Claude Opus 4.7 (xhigh) | Anthropic | 56.7% | $12.76 | 161 |
| 8 | Gemini 3 Flash | 54.1% | $0.58 | 114 | |
| 9 | Gemini 3.1 Pro | 53.2% | $0.96 | 77 | |
| 10 | Claude Haiku 4.5 | Anthropic | 44.5% | $0.55 | 72 |
| 11 | GPT 5 mini | OpenAI | 23.8% | $0.03 | 23 |
| 12 | GPT 5.4 mini | OpenAI | 18.6% | $0.02 | 8 |
| 13 | GPT 5.4 | OpenAI | 3.1% | $0.41 | 16 |