1114generated behavioral tests
93.1%official best score
89.9%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 | 89.9% | ok | 1002/1114 | $8.41 | 122 | 0.37h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 88.6% | ok | 987/1114 | n/a | 121 | 0.38h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 86.9% | ok | 968/1114 | n/a | 66 | 0.24h | 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
89.9% from 1002/1114 ProgramBench-scored tests. Raw public eval statuses: failure: 116, passed: 1140.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
e1213b502534eval.tests.test_edge_cases.test_negative_count_random (failure)e1213b502534eval.tests.test_random_distinct.test_random_invalid_strategy (failure)b3978b338c73tests.test_additional_errors.test_distinct_with_fixed_colors_exceeding_count (failure)b3978b338c73tests.test_basic_invocation.test_help_subcommand (failure)b3978b338c73tests.test_distinct_command.test_distinct_verbose (failure)b3978b338c73tests.test_colorcheck_pick_commands.test_pick_help (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 | 93.1% | $10.61 | 87 |
| 2 | Claude Opus 4.7 (xhigh) | Anthropic | 89.7% | $30.49 | 286 |
| 3 | GPT 5.5 (high) | OpenAI | 85.3% | $5.06 | 48 |
| 4 | GPT 5.5 | OpenAI | 77.8% | $1.84 | 21 |
| 5 | Claude Opus 4.6 | Anthropic | 77.2% | $17.50 | 232 |
| 6 | Claude Opus 4.7 | Anthropic | 74.6% | $3.22 | 92 |
| 7 | GPT 5.4 | OpenAI | 50.6% | $0.22 | 7 |
| 8 | Claude Haiku 4.5 | Anthropic | 35.4% | $1.22 | 179 |
| 9 | Gemini 3.1 Pro | 27.0% | $1.62 | 73 | |
| 10 | GPT 5 mini | OpenAI | 18.7% | $0.04 | 32 |
| 11 | Gemini 3 Flash | 17.8% | $0.38 | 82 | |
| 12 | GPT 5.4 mini | OpenAI | 9.4% | $0.04 | 21 |
| 13 | Claude Sonnet 4.6 | Anthropic | 0.0% | $6.72 | 321 |