603generated behavioral tests
71.0%official best score
64.2%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) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 64.2% | ok | 387/603 | n/a | 71 | 0.30h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 62.9% | ok | 379/603 | $4.71 | 89 | 0.22h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 55.4% | ok | 334/603 | n/a | 78 | 0.17h | 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) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260517T094700Z
62.9% from 379/603 ProgramBench-scored tests. Raw public eval statuses: failure: 240, passed: 510, skipped: 2.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
375015dacd58eval.tests.test_complex_scenarios.test_empty_directory (failure)375015dacd58eval.tests.test_basic_invocation.test_invalid_file (failure)375015dacd58eval.tests.test_edge_cases.test_function_with_context_parameter (failure)375015dacd58eval.tests.test_input_handling.test_file_with_syntax_errors (failure)375015dacd58eval.tests.test_templates.test_template_dir_precedence_over_template (failure)375015dacd58eval.tests.test_complex_scenarios.test_directory_with_no_go_files (failure)
GPT 5.5 (xhigh) · 20260518T232006Z
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 | 71.0% | $4.40 | 48 |
| 2 | GPT 5.5 (xhigh) | OpenAI | 62.2% | $7.99 | 102 |
| 3 | Claude Opus 4.6 | Anthropic | 61.9% | $14.13 | 274 |
| 4 | Claude Opus 4.7 | Anthropic | 61.0% | $4.87 | 105 |
| 5 | Claude Opus 4.7 (xhigh) | Anthropic | 52.6% | $15.81 | 190 |
| 6 | Claude Sonnet 4.6 | Anthropic | 52.2% | $31.36 | 542 |
| 7 | GPT 5.5 | OpenAI | 50.9% | $1.57 | 25 |
| 8 | GPT 5.4 | OpenAI | 45.1% | $0.59 | 16 |
| 9 | Gemini 3 Flash | 36.7% | $0.53 | 117 | |
| 10 | Claude Haiku 4.5 | Anthropic | 33.0% | $1.18 | 135 |
| 11 | GPT 5.4 mini | OpenAI | 25.0% | $0.04 | 12 |
| 12 | Gemini 3.1 Pro | 21.4% | $1.22 | 51 | |
| 13 | GPT 5 mini | OpenAI | 18.1% | $0.01 | 8 |