1931generated behavioral tests
58.4%official best score
42.0%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 | 42.0% | ok | 811/1931 | n/a | 49 | 0.17h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 23.5% | ok | 454/1931 | n/a | 118 | 0.37h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 16.4% | ok | 317/1931 | $5.85 | 94 | 0.29h | manifest · eval json · eval summary · usage audit |
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) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260517T094700Z
16.4% from 317/1931 ProgramBench-scored tests. Raw public eval statuses: failure: 2343, passed: 425.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, terminal rendering mismatch.
20ebc3204b43eval.tests.test_additional_features.test_combined_options (failure)20ebc3204b43eval.tests.test_animation.test_duration_numeric (failure)20ebc3204b43eval.tests.test_additional_features.test_format_and_color_combination (failure)20ebc3204b43eval.tests.test_additional_features.test_no_preprocess_with_high_colors (failure)20ebc3204b43eval.tests.test_additional_features.test_relative_positioning (failure)20ebc3204b43eval.tests.test_additional_features.test_polite_mode_combinations (failure)
Official ProgramBench Results by Model
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 4.6 | Anthropic | 58.4% | $50.80 | 532 |
| 2 | Claude Haiku 4.5 | Anthropic | 50.2% | $0.82 | 136 |
| 3 | Gemini 3 Flash | 47.1% | $0.17 | 65 | |
| 4 | Claude Opus 4.6 | Anthropic | 33.5% | $14.64 | 252 |
| 5 | GPT 5.5 (xhigh) | OpenAI | 28.0% | $8.49 | 77 |
| 6 | GPT 5.5 (high) | OpenAI | 21.4% | $3.78 | 44 |
| 7 | Claude Opus 4.7 (xhigh) | Anthropic | 18.7% | $7.73 | 146 |
| 8 | GPT 5.5 | OpenAI | 16.6% | $1.74 | 22 |
| 9 | Claude Opus 4.7 | Anthropic | 16.1% | $1.45 | 42 |
| 10 | GPT 5.4 | OpenAI | 8.7% | $0.16 | 7 |
| 11 | GPT 5 mini | OpenAI | 4.6% | $0.01 | 6 |
| 12 | GPT 5.4 mini | OpenAI | 3.5% | $0.04 | 9 |
| 13 | Gemini 3.1 Pro | 0.0% | $1.65 | 92 |