1994generated behavioral tests
84.1%official best score
81.9%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 | 81.9% | ok | 1634/1994 | $8.00 | 108 | 0.29h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 79.7% | ok | 1589/1994 | n/a | 89 | 0.26h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 74.8% | ok | 1491/1994 | n/a | 117 | 0.29h | 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
81.9% from 1634/1994 ProgramBench-scored tests. Raw public eval statuses: failure: 396, passed: 2139, skipped: 3.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
f177e1a6ce9eeval.tests.test_additional_coverage.test_generate_zsh (failure)f177e1a6ce9eeval.tests.test_additional_features.test_debug_output (failure)f177e1a6ce9eeval.tests.test_boost_to_50.test_config_file_with_debug (failure)f177e1a6ce9eeval.tests.test_boost_to_50.test_no_crlf (failure)f177e1a6ce9eeval.tests.test_additional_features.test_no_context_separator (failure)f177e1a6ce9eeval.tests.test_comprehensive_flags.test_flag_no_ignore_variations (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 | 84.1% | $17.56 | 121 |
| 2 | Claude Sonnet 4.6 | Anthropic | 79.7% | $19.39 | 508 |
| 3 | GPT 5.5 (high) | OpenAI | 70.5% | $3.00 | 28 |
| 4 | GPT 5.5 | OpenAI | 65.9% | $1.27 | 18 |
| 5 | Claude Opus 4.7 (xhigh) | Anthropic | 62.5% | $6.75 | 168 |
| 6 | Claude Opus 4.7 | Anthropic | 40.7% | $2.47 | 75 |
| 7 | GPT 5.4 | OpenAI | 35.5% | $0.26 | 7 |
| 8 | Gemini 3 Flash | 35.1% | $0.25 | 77 | |
| 9 | Gemini 3.1 Pro | 29.7% | $1.35 | 103 | |
| 10 | GPT 5 mini | OpenAI | 27.7% | $0.03 | 18 |
| 11 | Claude Haiku 4.5 | Anthropic | 23.7% | $0.88 | 118 |
| 12 | GPT 5.4 mini | OpenAI | 0.2% | $0.01 | 11 |
| 13 | Claude Opus 4.6 | Anthropic | 0.0% | $2.74 | 261 |