391generated behavioral tests
90.8%official best score
85.2%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) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 85.2% | ok | 333/391 | n/a | 79 | 0.26h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 82.6% | ok | 323/391 | n/a | 64 | 0.20h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 79.8% | ok | 312/391 | $5.09 | 64 | 0.28h | manifest · eval json · eval summary · usage audit |
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) · 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.
GPT 5.5 (xhigh) · 20260517T094700Z
79.8% from 312/391 ProgramBench-scored tests. Raw public eval statuses: failure: 83, passed: 336.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
8384425dbf58eval.tests.test_basic.TestInputHandling.test_empty_file_path (failure)8384425dbf58eval.tests.test_basic.TestOutputGeneration.test_output_same_as_input_error (failure)79af22e6a745eval.tests.test_error_handling.test_output_file_same_as_input (failure)79af22e6a745eval.tests.test_error_handling.test_empty_wsdl_file (failure)79af22e6a745eval.tests.test_error_handling.test_treat_last_arg_as_wsdl (failure)79af22e6a745eval.tests.test_error_handling.test_invalid_xml_but_looks_like_wsdl (failure)
Official ProgramBench Results by Model
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | GPT 5.5 (xhigh) | OpenAI | 90.8% | $17.46 | 117 |
| 2 | Claude Opus 4.7 (xhigh) | Anthropic | 89.5% | $21.55 | 224 |
| 3 | Claude Opus 4.7 | Anthropic | 86.4% | $19.71 | 260 |
| 4 | GPT 5.5 (high) | OpenAI | 85.7% | $7.43 | 73 |
| 5 | GPT 5.5 | OpenAI | 80.6% | $1.60 | 28 |
| 6 | Gemini 3.1 Pro | 69.3% | $1.04 | 82 | |
| 7 | Claude Haiku 4.5 | Anthropic | 63.2% | $1.02 | 155 |
| 8 | Claude Sonnet 4.6 | Anthropic | 62.4% | $60.18 | 801 |
| 9 | Claude Opus 4.6 | Anthropic | 61.6% | $20.18 | 296 |
| 10 | Gemini 3 Flash | 30.7% | $0.27 | 62 | |
| 11 | GPT 5.4 mini | OpenAI | 26.6% | $0.02 | 13 |
| 12 | GPT 5.4 | OpenAI | 11.8% | $0.38 | 10 |
| 13 | GPT 5 mini | OpenAI | 11.5% | $0.01 | 9 |