1300generated behavioral tests
93.9%official best score
94.7%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 | 94.7% | ok | 1231/1300 | $8.56 | 127 | 0.44h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 64.6% | ok | 840/1300 | n/a | 63 | 0.21h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 55.8% | ok | 726/1300 | n/a | 66 | 0.25h | 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
94.7% from 1231/1300 ProgramBench-scored tests. Raw public eval statuses: failure: 244, passed: 1473, skipped: 5.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch, interactive terminal behavior mismatch.
c25f375009bftests.test_additional_edge_cases.test_invalid_color_setting (failure)c25f375009bftests.test_color_output.test_help_color_always (failure)c25f375009bftests.test_color_output.test_version_color_always (failure)c25f375009bftests.test_color_output.test_help_colour_always_british (failure)c25f375009bftests.test_error_handling.test_https_without_nameserver (failure)c25f375009bftests.test_color_output.test_version_colour_always_british (failure)
GPT 5.5 (xhigh) · 20260518T232006Z
Public eval evidence has not been exported for this row yet.
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 | 93.9% | $8.10 | 68 |
| 2 | GPT 5.5 (high) | OpenAI | 91.2% | $3.43 | 41 |
| 3 | Claude Opus 4.7 (xhigh) | Anthropic | 90.5% | $15.42 | 222 |
| 4 | Claude Opus 4.6 | Anthropic | 84.2% | $11.72 | 273 |
| 5 | Claude Opus 4.7 | Anthropic | 80.8% | $5.47 | 144 |
| 6 | Claude Sonnet 4.6 | Anthropic | 76.5% | $7.18 | 249 |
| 7 | GPT 5.5 | OpenAI | 75.4% | $1.32 | 20 |
| 8 | Gemini 3.1 Pro | 70.8% | $1.14 | 99 | |
| 9 | Gemini 3 Flash | 69.0% | $0.30 | 94 | |
| 10 | Claude Haiku 4.5 | Anthropic | 51.7% | $0.74 | 112 |
| 11 | GPT 5.4 | OpenAI | 28.8% | $0.17 | 7 |
| 12 | GPT 5.4 mini | OpenAI | 17.8% | $0.03 | 7 |
| 13 | GPT 5 mini | OpenAI | 10.2% | $0.02 | 17 |