1978generated behavioral tests
47.4%official best score
46.1%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 | 46.1% | ok | 911/1978 | $4.28 | 82 | 0.21h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 36.9% | ok | 729/1978 | n/a | 84 | 0.23h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 33.9% | ok | 670/1978 | n/a | 65 | 0.25h | 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
46.1% from 911/1978 ProgramBench-scored tests. Raw public eval statuses: error: 1, failure: 1105, passed: 1420, skipped: 2.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
69f0eeebd3b8eval.tests.test_basic_invocation.test_no_arguments (failure)69f0eeebd3b8eval.tests.test_coverage_gaps.test_multiple_stdin_inputs (failure)69f0eeebd3b8eval.tests.test_coverage_gaps.test_yul_file_without_parse_yul_flag (failure)69f0eeebd3b8eval.tests.test_coverage_gaps.test_diagnostic_width_option (failure)69f0eeebd3b8eval.tests.test_coverage_gaps.test_error_format_json_with_syntax_error (failure)69f0eeebd3b8eval.tests.test_coverage_gaps.test_pretty_json_err_with_error (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 | 47.4% | $8.25 | 79 |
| 2 | GPT 5.5 (high) | OpenAI | 46.4% | $3.03 | 36 |
| 3 | Claude Opus 4.6 | Anthropic | 43.3% | $14.23 | 245 |
| 4 | Claude Opus 4.7 | Anthropic | 43.0% | $3.30 | 98 |
| 5 | Claude Sonnet 4.6 | Anthropic | 42.9% | $15.75 | 367 |
| 6 | GPT 5.5 | OpenAI | 40.3% | $1.50 | 24 |
| 7 | Claude Opus 4.7 (xhigh) | Anthropic | 39.1% | $5.84 | 105 |
| 8 | Claude Haiku 4.5 | Anthropic | 32.5% | $0.80 | 103 |
| 9 | Gemini 3.1 Pro | 32.3% | $2.40 | 146 | |
| 10 | Gemini 3 Flash | 22.3% | $0.36 | 100 | |
| 11 | GPT 5.4 | OpenAI | 11.5% | $0.16 | 8 |
| 12 | GPT 5 mini | OpenAI | 3.3% | $0.01 | 11 |
| 13 | GPT 5.4 mini | OpenAI | 0.0% | $0.02 | 8 |