479generated behavioral tests
81.6%official best score
67.2%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 | 67.2% | ok | 322/479 | $3.97 | 60 | 0.23h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 57.0% | ok | 273/479 | n/a | 46 | 0.18h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 55.5% | ok | 266/479 | n/a | 61 | 0.18h | 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
67.2% from 322/479 ProgramBench-scored tests. Raw public eval statuses: error: 1, failure: 169, passed: 389, skipped: 5.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, interactive terminal behavior mismatch, terminal rendering mismatch.
4e50f150bcfaeval.tests.test_edge_cases.test_dash_dash_help (failure)4e50f150bcfaeval.tests.test_more_parquet_files.test_delta_byte_array (failure)4e50f150bcfaeval.tests.test_more_parquet_files.test_fixed_length_decimal_parquet (failure)4e50f150bcfaeval.tests.test_more_parquet_files.test_fixed_length_decimal_legacy (failure)4e50f150bcfaeval.tests.test_parquet_formats.test_alltypes_dictionary (failure)4e50f150bcfaeval.tests.test_parquet_formats.test_alltypes_tiny_pages (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 | 81.6% | $6.37 | 55 |
| 2 | GPT 5.5 (high) | OpenAI | 64.1% | $4.43 | 39 |
| 3 | Claude Opus 4.7 (xhigh) | Anthropic | 64.1% | $4.31 | 112 |
| 4 | GPT 5.5 | OpenAI | 60.3% | $1.66 | 24 |
| 5 | Claude Opus 4.7 | Anthropic | 58.9% | $1.25 | 40 |
| 6 | GPT 5.4 | OpenAI | 48.4% | $0.21 | 14 |
| 7 | Claude Opus 4.6 | Anthropic | 45.5% | $14.19 | 203 |
| 8 | Claude Haiku 4.5 | Anthropic | 34.4% | $1.16 | 181 |
| 9 | GPT 5 mini | OpenAI | 13.8% | $0.02 | 17 |
| 10 | Claude Sonnet 4.6 | Anthropic | 13.2% | $91.59 | 720 |
| 11 | Gemini 3.1 Pro | 4.2% | $0.82 | 46 | |
| 12 | Gemini 3 Flash | 0.0% | $0.41 | 105 | |
| 13 | GPT 5.4 mini | OpenAI | 0.0% | $0.02 | 18 |