575generated behavioral tests
96.2%official best score
94.1%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) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 94.1% | ok | 541/575 | n/a | 61 | 0.21h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 93.9% | ok | 540/575 | $4.44 | 77 | 0.24h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 89.9% | ok | 517/575 | n/a | 72 | 0.23h | 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
93.9% from 540/575 ProgramBench-scored tests. Raw public eval statuses: failure: 35, passed: 579, skipped: 1.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
0b8272ce2673tests.test_verbosity.test_quiet_and_verbose_conflict (failure)f56b4714726deval.tests.test_additional_formats.test_gif_to_webp (failure)f56b4714726deval.tests.test_minimum_savings.test_min_savings_allows_writing (failure)f56b4714726deval.tests.test_output_control.test_quiet_overrides_verbose (failure)f56b4714726deval.tests.test_output_verbosity.test_quiet_and_verbose_mutually_exclusive (failure)8b3562091b25tests.test_compression.test_keep_structure_without_recursive_error (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 (xhigh) | OpenAI | 96.2% | $7.95 | 71 |
| 2 | GPT 5.5 (high) | OpenAI | 92.5% | $5.06 | 60 |
| 3 | Claude Opus 4.6 | Anthropic | 92.3% | $11.47 | 294 |
| 4 | Claude Opus 4.7 (xhigh) | Anthropic | 76.5% | $9.10 | 164 |
| 5 | GPT 5.5 | OpenAI | 76.3% | $1.35 | 23 |
| 6 | Claude Opus 4.7 | Anthropic | 73.6% | $1.89 | 57 |
| 7 | Gemini 3.1 Pro | 58.1% | $1.54 | 95 | |
| 8 | Claude Haiku 4.5 | Anthropic | 57.6% | $0.86 | 135 |
| 9 | Gemini 3 Flash | 49.4% | $0.40 | 139 | |
| 10 | GPT 5.4 mini | OpenAI | 17.2% | $0.04 | 28 |
| 11 | GPT 5 mini | OpenAI | 12.5% | $0.02 | 12 |
| 12 | Claude Sonnet 4.6 | Anthropic | 5.7% | $8.73 | 350 |
| 13 | GPT 5.4 | OpenAI | 1.6% | $0.61 | 17 |