767generated behavioral tests
81.7%official best score
2.1%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) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 2.1% | ok | 16/767 | $5.98 | 109 | 0.30h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 2.1% | ok | 16/767 | n/a | 34 | 0.12h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 2.1% | ok | 16/767 | n/a | 53 | 0.18h | 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
2.1% from 16/767 ProgramBench-scored tests. Raw public eval statuses: failure: 795, passed: 29, skipped: 1.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
3d13bf8d9dfbeval.tests.test_additional_coverage.test_yaml_to_json_with_floatstrings_disabled (failure)3d13bf8d9dfbeval.tests.test_additional_coverage.test_toml_array_of_tables (failure)3d13bf8d9dfbeval.tests.test_additional_coverage.test_toml_inline_tables (failure)3d13bf8d9dfbeval.tests.test_edge_cases.test_numeric_string_keys (failure)3d13bf8d9dfbeval.tests.test_basic.test_help_flag (failure)3d13bf8d9dfbeval.tests.test_additional_coverage.test_toml_to_json_with_indent_and_nonan (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 | GPT 5.5 (xhigh) | OpenAI | 81.7% | $5.69 | 52 |
| 2 | Claude Opus 4.7 (xhigh) | Anthropic | 78.7% | $16.06 | 196 |
| 3 | Claude Opus 4.7 | Anthropic | 74.4% | $4.14 | 88 |
| 4 | GPT 5.5 (high) | OpenAI | 73.5% | $2.48 | 34 |
| 5 | GPT 5.5 | OpenAI | 66.5% | $0.79 | 15 |
| 6 | Gemini 3.1 Pro | 52.4% | $1.40 | 85 | |
| 7 | GPT 5.4 | OpenAI | 50.2% | $0.32 | 13 |
| 8 | Claude Haiku 4.5 | Anthropic | 46.3% | $1.61 | 184 |
| 9 | GPT 5 mini | OpenAI | 31.9% | $0.02 | 13 |
| 10 | Claude Sonnet 4.6 | Anthropic | 2.0% | $25.56 | 730 |
| 11 | GPT 5.4 mini | OpenAI | 2.0% | $0.02 | 8 |
| 12 | Claude Opus 4.6 | Anthropic | 0.0% | $2.92 | 293 |
| 13 | Gemini 3 Flash | 0.0% | $0.23 | 112 |