5650generated behavioral tests
16.8%official best score
16.3%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 | 16.3% | ok | 919/5650 | $6.77 | 121 | 0.29h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 8.2% | ok | 466/5650 | n/a | 40 | 0.14h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 8.1% | ok | 457/5650 | n/a | 88 | 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) · 20260517T094700Z
16.3% from 919/5650 ProgramBench-scored tests. Raw public eval statuses: failure: 6730, passed: 1093, skipped: 1134.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch, interactive terminal behavior mismatch.
3466718f282beval.tests.test_duckdb_cli.TestCommandLineFlags.test_echo_flag (failure)3466718f282beval.tests.test_duckdb_cli.TestCommandLineFlags.test_readonly_flag_allows_reads (failure)3466718f282beval.tests.test_duckdb_cli.TestOutputModes.test_default_mode_is_duckbox (failure)3466718f282beval.tests.test_duckdb_cli.TestCommandLineFlags.test_storage_version_valid_latest (failure)3466718f282beval.tests.test_duckdb_cli.TestCommandLineFlags.test_batch_flag (failure)3466718f282beval.tests.test_duckdb_cli.TestBasicInvocation.test_no_stdin_exits_cleanly (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 | 16.8% | $7.51 | 77 |
| 2 | GPT 5.5 (high) | OpenAI | 12.8% | $3.92 | 35 |
| 3 | Claude Opus 4.6 | Anthropic | 12.4% | $17.56 | 441 |
| 4 | GPT 5.5 | OpenAI | 11.9% | $0.99 | 15 |
| 5 | GPT 5.4 | OpenAI | 10.9% | $0.43 | 13 |
| 6 | Gemini 3.1 Pro | 6.2% | $1.01 | 88 | |
| 7 | Gemini 3 Flash | 4.7% | $0.49 | 118 | |
| 8 | GPT 5.4 mini | OpenAI | 4.3% | $0.03 | 57 |
| 9 | Claude Opus 4.7 (xhigh) | Anthropic | 4.1% | $0.70 | 32 |
| 10 | Claude Opus 4.7 | Anthropic | 4.0% | $1.41 | 62 |
| 11 | GPT 5 mini | OpenAI | 4.0% | $0.03 | 14 |
| 12 | Claude Haiku 4.5 | Anthropic | 3.8% | $0.86 | 121 |
| 13 | Claude Sonnet 4.6 | Anthropic | 0.0% | $8.24 | 288 |