1963generated behavioral tests
83.7%official best score
79.3%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 | 79.3% | ok | 1556/1963 | $5.45 | 82 | 0.31h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 70.8% | ok | 1389/1963 | n/a | 107 | 0.25h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 69.1% | ok | 1357/1963 | n/a | 67 | 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
79.3% from 1556/1963 ProgramBench-scored tests. Raw public eval statuses: failure: 421, passed: 1856, skipped: 12.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
90713bdac2ddeval.tests.test_coverage_50_boost.test_cd_dash_previous (failure)90713bdac2ddeval.tests.test_control_structures.test_case_statement (failure)90713bdac2ddeval.tests.test_comprehensive.test_arithmetic_negative (failure)90713bdac2ddeval.tests.test_conditionals.test_double_bracket_or (failure)90713bdac2ddeval.tests.test_coverage_50_boost.test_source_nonexistent (failure)90713bdac2ddeval.tests.test_control_structures.test_c_style_for_loop (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 | Gemini 3.1 Pro | 83.7% | $0.92 | 64 | |
| 2 | GPT 5.5 (xhigh) | OpenAI | 81.0% | $9.98 | 81 |
| 3 | Claude Sonnet 4.6 | Anthropic | 80.6% | $27.07 | 496 |
| 4 | Claude Opus 4.6 | Anthropic | 76.9% | $13.22 | 304 |
| 5 | GPT 5 mini | OpenAI | 76.1% | $0.01 | 10 |
| 6 | Claude Opus 4.7 (xhigh) | Anthropic | 71.3% | $5.86 | 116 |
| 7 | Claude Opus 4.7 | Anthropic | 66.1% | $3.70 | 92 |
| 8 | GPT 5.5 | OpenAI | 52.9% | $1.33 | 23 |
| 9 | GPT 5.4 | OpenAI | 50.2% | $0.19 | 8 |
| 10 | Claude Haiku 4.5 | Anthropic | 36.8% | $1.16 | 148 |
| 11 | Gemini 3 Flash | 31.3% | $0.28 | 122 | |
| 12 | GPT 5.4 mini | OpenAI | 0.0% | $0.02 | 14 |
| 13 | GPT 5.5 (high) | OpenAI | n/a | $2.79 | 42 |