713generated behavioral tests
67.7%official best score
63.8%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 | 63.8% | ok | 455/713 | $4.25 | 77 | 0.21h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 49.1% | ok | 350/713 | n/a | 77 | 0.27h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 48.9% | ok | 349/713 | n/a | 69 | 0.24h | 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
63.8% from 455/713 ProgramBench-scored tests. Raw public eval statuses: failure: 259, passed: 518.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
4da2290d51e2tests.test_cookies.test_cookie_file_nonexistent (failure)1f0786a07bddeval.tests.test_cli_operations.test_isolate_with_no_css_combined (failure)1f0786a07bddeval.tests.test_cli_operations.test_no_js_flag (failure)1f0786a07bddeval.tests.test_cli_operations.test_timeout_flag (failure)1f0786a07bddeval.tests.test_cli_operations.test_output_to_stdout_explicit_dash (failure)1f0786a07bddeval.tests.test_cli_operations.test_stdin_with_malformed_html (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 | 67.7% | $6.59 | 61 |
| 2 | GPT 5.5 (high) | OpenAI | 60.6% | $4.38 | 57 |
| 3 | Claude Opus 4.7 (xhigh) | Anthropic | 56.2% | $6.26 | 113 |
| 4 | GPT 5.5 | OpenAI | 51.8% | $1.00 | 19 |
| 5 | Gemini 3 Flash | 51.2% | $0.33 | 72 | |
| 6 | Gemini 3.1 Pro | 38.7% | $0.95 | 77 | |
| 7 | Claude Haiku 4.5 | Anthropic | 36.7% | $0.89 | 121 |
| 8 | GPT 5.4 mini | OpenAI | 33.0% | $0.03 | 8 |
| 9 | GPT 5.4 | OpenAI | 25.8% | $0.25 | 8 |
| 10 | Claude Opus 4.7 | Anthropic | 2.4% | $3.80 | 118 |
| 11 | Claude Sonnet 4.6 | Anthropic | 2.4% | $14.72 | 406 |
| 12 | Claude Opus 4.6 | Anthropic | 1.3% | $12.81 | 305 |
| 13 | GPT 5 mini | OpenAI | 1.3% | $0.03 | 16 |