2038generated behavioral tests
68.8%official best score
69.6%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 | 69.6% | ok | 1419/2038 | $4.54 | 77 | 0.23h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 63.7% | ok | 1299/2038 | n/a | 50 | 0.17h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 61.3% | ok | 1250/2038 | n/a | 45 | 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
69.6% from 1419/2038 ProgramBench-scored tests. Raw public eval statuses: failure: 656, passed: 1697, skipped: 16.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
be378a85aca3eval.tests.test_02_compression.test_compress_explicit_flag (failure)be378a85aca3eval.tests.test_07_dictionaries.test_train_dictionary_basic (failure)be378a85aca3eval.tests.test_14_sparse.test_sparse_mode_with_file (failure)be378a85aca3eval.tests.test_17_additional_options.test_no_dictID_option (failure)be378a85aca3eval.tests.test_17_additional_options.test_no_mmap_dict_option (failure)be378a85aca3eval.tests.test_18_error_handling.test_decompress_truncated_data (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 Sonnet 4.6 | Anthropic | 68.8% | $15.29 | 375 |
| 2 | GPT 5.5 (xhigh) | OpenAI | 68.7% | $5.13 | 64 |
| 3 | GPT 5.5 (high) | OpenAI | 62.4% | $2.54 | 34 |
| 4 | Claude Opus 4.7 (xhigh) | Anthropic | 61.6% | $9.84 | 142 |
| 5 | GPT 5.5 | OpenAI | 58.4% | $1.24 | 22 |
| 6 | Claude Opus 4.7 | Anthropic | 55.0% | $3.68 | 104 |
| 7 | Gemini 3 Flash | 50.8% | $0.21 | 73 | |
| 8 | Gemini 3.1 Pro | 44.9% | $1.11 | 98 | |
| 9 | GPT 5.4 | OpenAI | 35.3% | $0.21 | 11 |
| 10 | Claude Haiku 4.5 | Anthropic | 33.9% | $0.84 | 129 |
| 11 | GPT 5.4 mini | OpenAI | 7.7% | $0.02 | 5 |
| 12 | GPT 5 mini | OpenAI | 7.0% | $0.02 | 7 |
| 13 | Claude Opus 4.6 | Anthropic | 0.0% | $1.41 | 189 |