14288generated behavioral tests
13.2%official best score
11.9%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 | 11.9% | ok | 1696/14288 | $8.04 | 106 | 0.39h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 8.8% | ok | 1254/14288 | n/a | 64 | 0.24h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 8.5% | ok | 1218/14288 | n/a | 82 | 0.32h | 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
11.9% from 1696/14288 ProgramBench-scored tests. Raw public eval statuses: error: 543, failure: 18060, passed: 1780, skipped: 147.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
c91c8a170be8eval.tests.test_closures_and_functions.TestFunctionFeatures.test_spread_operator (failure)c91c8a170be8eval.tests.test_advanced_features.TestErrorHandling.test_set_error_handler (failure)c91c8a170be8eval.tests.test_advanced_php_features.TestCallbacksAndCallables.test_usort_with_callback (failure)c91c8a170be8eval.tests.test_additional_features.TestRepeatFlag.test_repeat_execution (failure)c91c8a170be8eval.tests.test_comprehensive_features.TestTraits.test_trait_usage (failure)c91c8a170be8eval.tests.test_comprehensive_math_date.TestDateTimeFunctions.test_localtime (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 | 13.2% | $12.89 | 115 |
| 2 | GPT 5.5 (high) | OpenAI | 8.0% | $3.23 | 36 |
| 3 | Claude Haiku 4.5 | Anthropic | 4.8% | $0.52 | 119 |
| 4 | Claude Opus 4.7 (xhigh) | Anthropic | 3.2% | $3.26 | 73 |
| 5 | GPT 5.4 | OpenAI | 2.2% | $0.24 | 7 |
| 6 | Gemini 3.1 Pro | 1.7% | $1.33 | 109 | |
| 7 | GPT 5.4 mini | OpenAI | 1.1% | $0.03 | 13 |
| 8 | GPT 5 mini | OpenAI | 0.9% | $0.04 | 14 |
| 9 | Claude Opus 4.7 | Anthropic | 0.9% | $0.25 | 14 |
| 10 | Gemini 3 Flash | 0.6% | $0.22 | 65 | |
| 11 | Claude Opus 4.6 | Anthropic | 0.0% | $2.94 | 205 |
| 12 | Claude Sonnet 4.6 | Anthropic | 0.0% | $7.38 | 326 |
| 13 | GPT 5.5 | OpenAI | n/a | $1.47 | 16 |