1748generated behavioral tests
88.0%official best score
78.9%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) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 78.9% | ok | 1379/1748 | n/a | 61 | 0.20h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 78.3% | ok | 1369/1748 | $5.75 | 90 | 0.27h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 77.3% | ok | 1351/1748 | n/a | 82 | 0.27h | 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) · 20260518T232006Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260517T094700Z
78.3% from 1369/1748 ProgramBench-scored tests. Raw public eval statuses: error: 1, failure: 731, passed: 2366, skipped: 6.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch, interactive terminal behavior mismatch.
01eed97aa21ctests.test_completion.TestCompletion.test_completion_fish (failure)01eed97aa21ctests.test_cmd_helper_functions.TestCompletionHelpers.test_completion_fish_structure (failure)01eed97aa21ctests.test_coverage_targeted.TestValidationEdgeCases.test_empty_service_name (failure)01eed97aa21ctests.test_env_command.TestEnvCommand.test_env_invalid_service (failure)01eed97aa21ctests.test_env_command.TestEnvCommandEdgeCases.test_env_empty_service_name (failure)01eed97aa21ctests.test_edge_cases.TestSpecialCharacters.test_empty_string_service (failure)
GPT 5.5 (xhigh) · 20260519T215046Z
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 | 88.0% | $15.51 | 99 |
| 2 | Claude Opus 4.6 | Anthropic | 82.0% | $11.99 | 232 |
| 3 | Claude Opus 4.7 (xhigh) | Anthropic | 79.3% | $7.80 | 140 |
| 4 | GPT 5.5 (high) | OpenAI | 79.0% | $3.86 | 46 |
| 5 | GPT 5.5 | OpenAI | 71.3% | $1.20 | 18 |
| 6 | Claude Sonnet 4.6 | Anthropic | 62.1% | $14.17 | 359 |
| 7 | GPT 5.4 | OpenAI | 56.2% | $0.24 | 9 |
| 8 | Claude Haiku 4.5 | Anthropic | 43.0% | $0.78 | 117 |
| 9 | Gemini 3 Flash | 40.3% | $0.21 | 77 | |
| 10 | GPT 5 mini | OpenAI | 20.9% | $0.04 | 24 |
| 11 | Gemini 3.1 Pro | 19.9% | $1.34 | 66 | |
| 12 | GPT 5.4 mini | OpenAI | 14.5% | $0.03 | 9 |
| 13 | Claude Opus 4.7 | Anthropic | 9.4% | $3.95 | 92 |