373generated behavioral tests
92.8%official best score
87.4%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 | 87.4% | ok | 326/373 | $4.82 | 85 | 0.23h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 87.1% | ok | 325/373 | n/a | 69 | 0.21h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 81.2% | ok | 303/373 | n/a | 60 | 0.21h | 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
87.4% from 326/373 ProgramBench-scored tests. Raw public eval statuses: failure: 48, passed: 402.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
6388a2824327eval.tests.test_dupl.test_conflicting_html_plumbing_flags (failure)f25ebcd2692etests.test_dupl.test_html_and_plumbing_conflict (failure)9dea7bde779btests.test_errors.test_non_go_text_file (failure)9dea7bde779btests.test_errors.test_malformed_go_file (failure)9dea7bde779btests.test_errors.test_binary_file (failure)9dea7bde779btests.test_go_constructs.test_parse_error_handling (failure)
GPT 5.5 (xhigh) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
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 (xhigh) | OpenAI | 92.8% | $6.26 | 67 |
| 2 | Claude Opus 4.7 (xhigh) | Anthropic | 87.1% | $10.07 | 123 |
| 3 | GPT 5.5 (high) | OpenAI | 86.9% | $2.56 | 34 |
| 4 | Claude Opus 4.7 | Anthropic | 85.0% | $4.93 | 88 |
| 5 | GPT 5.5 | OpenAI | 84.2% | $1.02 | 15 |
| 6 | Claude Opus 4.6 | Anthropic | 79.6% | $13.85 | 215 |
| 7 | Claude Sonnet 4.6 | Anthropic | 71.6% | $45.63 | 514 |
| 8 | Gemini 3 Flash | 66.5% | $0.26 | 85 | |
| 9 | GPT 5.4 | OpenAI | 58.4% | $0.45 | 13 |
| 10 | GPT 5.4 mini | OpenAI | 57.1% | $0.03 | 9 |
| 11 | Gemini 3.1 Pro | 48.0% | $1.30 | 71 | |
| 12 | Claude Haiku 4.5 | Anthropic | 38.3% | $0.69 | 111 |
| 13 | GPT 5 mini | OpenAI | 35.4% | $0.02 | 13 |