997generated behavioral tests
93.3%official best score
92.8%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 | 92.8% | ok | 925/997 | $3.14 | 55 | 0.21h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 87.3% | ok | 870/997 | n/a | 66 | 0.20h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 79.0% | ok | 788/997 | n/a | 49 | 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) · 20260517T094700Z
92.8% from 925/997 ProgramBench-scored tests. Raw public eval statuses: failure: 72, passed: 980, skipped: 1.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
10d3210b1675tests.test_weak_fixes.test_encode_function_with_invalid_json_abi_shows_error (failure)d9fde585fa84eval.tests.test_decode_params.test_decode_params_invalid_type_fallback (failure)d9fde585fa84eval.tests.test_encode_function.test_encode_function_extra_params_ignored (failure)d9fde585fa84eval.tests.test_error_paths.test_encode_uint_overflow (failure)13dc3ceeb33dtests.test_decode_log.test_error_topic_wrong_length (failure)13dc3ceeb33dtests.test_abi_edge_cases.test_tuple_param_duplicate_type (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 | Claude Opus 4.7 (xhigh) | Anthropic | 93.3% | $14.23 | 225 |
| 2 | Claude Opus 4.7 | Anthropic | 90.9% | $13.93 | 246 |
| 3 | Claude Sonnet 4.6 | Anthropic | 89.9% | $12.55 | 407 |
| 4 | GPT 5.5 (xhigh) | OpenAI | 88.7% | $6.65 | 73 |
| 5 | GPT 5.5 (high) | OpenAI | 86.1% | $4.62 | 44 |
| 6 | Claude Opus 4.6 | Anthropic | 84.0% | $10.26 | 257 |
| 7 | GPT 5.5 | OpenAI | 74.3% | $0.98 | 15 |
| 8 | Claude Haiku 4.5 | Anthropic | 52.2% | $1.25 | 159 |
| 9 | GPT 5.4 | OpenAI | 49.3% | $0.32 | 12 |
| 10 | GPT 5 mini | OpenAI | 30.5% | $0.03 | 25 |
| 11 | Gemini 3.1 Pro | 7.8% | $2.23 | 160 | |
| 12 | Gemini 3 Flash | 7.8% | $0.23 | 89 | |
| 13 | GPT 5.4 mini | OpenAI | 0.0% | $0.03 | 14 |