807generated behavioral tests
60.3%official best score
59.9%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 | 59.9% | ok | 483/807 | $4.62 | 68 | 0.25h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 55.1% | ok | 445/807 | n/a | 37 | 0.15h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 51.4% | ok | 415/807 | n/a | 49 | 0.22h | 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
59.9% from 483/807 ProgramBench-scored tests. Raw public eval statuses: failure: 334, passed: 553.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
547860b2846eeval.tests.test_actual_rendering.test_render_with_cover_image (failure)547860b2846eeval.tests.test_actual_rendering.test_render_to_stdout (failure)547860b2846eeval.tests.test_create_book.test_create_book_with_filename (failure)547860b2846eeval.tests.test_create_book.test_create_book_file_already_exists (failure)547860b2846eeval.tests.test_output_formats.test_html_dir_format (failure)547860b2846eeval.tests.test_print_template.test_print_template_multiple_types (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 | Claude Sonnet 4.6 | Anthropic | 60.3% | $85.27 | 937 |
| 2 | GPT 5.5 (high) | OpenAI | 48.8% | $3.32 | 38 |
| 3 | Claude Opus 4.6 | Anthropic | 45.0% | $19.00 | 320 |
| 4 | GPT 5.5 | OpenAI | 43.2% | $1.04 | 15 |
| 5 | Claude Opus 4.7 (xhigh) | Anthropic | 29.4% | $13.25 | 195 |
| 6 | Gemini 3.1 Pro | 22.4% | $0.98 | 81 | |
| 7 | Claude Haiku 4.5 | Anthropic | 21.2% | $0.47 | 106 |
| 8 | GPT 5.4 | OpenAI | 20.6% | $0.34 | 11 |
| 9 | GPT 5 mini | OpenAI | 13.6% | $0.04 | 17 |
| 10 | Gemini 3 Flash | 12.8% | $0.45 | 92 | |
| 11 | GPT 5.4 mini | OpenAI | 11.5% | $0.03 | 9 |
| 12 | GPT 5.5 (xhigh) | OpenAI | 2.9% | $10.81 | 98 |
| 13 | Claude Opus 4.7 | Anthropic | 0.9% | $2.01 | 69 |