727generated behavioral tests
48.3%official best score
37.1%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 | 37.1% | ok | 270/727 | n/a | 88 | 0.21h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 33.8% | ok | 246/727 | $5.35 | 100 | 0.26h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 31.4% | ok | 228/727 | n/a | 67 | 0.26h | 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) · 20260518T232006Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260517T094700Z
33.8% from 246/727 ProgramBench-scored tests. Raw public eval statuses: failure: 514, passed: 372.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
836b90647508eval.tests.test_comment_directives.test_revive_disable_enable_directive (failure)836b90647508eval.tests.test_comment_directives.test_revive_disable_next_line_directive (failure)836b90647508eval.tests.test_comment_directives.test_directive_with_reason (failure)836b90647508eval.tests.test_comment_directives.test_disable_multiple_rules (failure)836b90647508eval.tests.test_comment_directives.test_disable_in_middle_of_file (failure)836b90647508eval.tests.test_exclusions.test_exclude_single_file (failure)
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 (high) | OpenAI | 48.3% | $3.70 | 56 |
| 2 | GPT 5.5 | OpenAI | 47.3% | $1.24 | 22 |
| 3 | Claude Opus 4.7 (xhigh) | Anthropic | 46.4% | $2.67 | 70 |
| 4 | Claude Opus 4.6 | Anthropic | 46.4% | $10.31 | 254 |
| 5 | GPT 5.5 (xhigh) | OpenAI | 45.9% | $6.53 | 64 |
| 6 | Claude Opus 4.7 | Anthropic | 43.1% | $2.52 | 62 |
| 7 | Claude Sonnet 4.6 | Anthropic | 34.1% | $17.26 | 476 |
| 8 | Claude Haiku 4.5 | Anthropic | 25.9% | $0.55 | 149 |
| 9 | Gemini 3 Flash | 25.0% | $0.24 | 61 | |
| 10 | Gemini 3.1 Pro | 24.8% | $1.12 | 75 | |
| 11 | GPT 5.4 | OpenAI | 21.3% | $0.38 | 15 |
| 12 | GPT 5.4 mini | OpenAI | 6.2% | $0.02 | 8 |
| 13 | GPT 5 mini | OpenAI | 4.8% | $0.01 | 8 |