445generated behavioral tests
93.0%official best score
71.2%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 | 71.2% | ok | 317/445 | n/a | 47 | 0.15h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 70.1% | ok | 312/445 | $2.56 | 54 | 0.15h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 69.7% | ok | 310/445 | n/a | 46 | 0.13h | 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
70.1% from 312/445 ProgramBench-scored tests. Raw public eval statuses: failure: 147, passed: 360.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, interactive terminal behavior mismatch, terminal rendering mismatch.
30e1ad7f2bf9eval.tests.test_tui.TestTUIResults.test_results_show_wpm (failure)30e1ad7f2bf9eval.tests.test_tui.TestTUIResults.test_results_show_options (failure)30e1ad7f2bf9eval.tests.test_tui.TestTUIInteraction.test_tui_typing_completes_test (failure)30e1ad7f2bf9eval.tests.test_tui.TestTUIResults.test_results_show_accuracy (failure)30e1ad7f2bf9eval.tests.test_tui.TestTUIResults.test_results_show_standard_deviation (failure)2a583613c55beval.tests.test_tui_interactive.test_tui_navigation_keys (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 | Claude Opus 4.7 (xhigh) | Anthropic | 93.0% | $18.61 | 257 |
| 2 | GPT 5.5 (xhigh) | OpenAI | 92.8% | $6.84 | 60 |
| 3 | GPT 5.5 (high) | OpenAI | 83.8% | $2.42 | 29 |
| 4 | GPT 5.5 | OpenAI | 83.4% | $1.04 | 19 |
| 5 | Claude Opus 4.6 | Anthropic | 82.2% | $18.03 | 328 |
| 6 | Claude Opus 4.7 | Anthropic | 63.1% | $4.85 | 126 |
| 7 | GPT 5.4 | OpenAI | 43.4% | $0.09 | 6 |
| 8 | Gemini 3.1 Pro | 42.5% | $1.31 | 89 | |
| 9 | Gemini 3 Flash | 40.2% | $0.32 | 96 | |
| 10 | GPT 5.4 mini | OpenAI | 29.9% | $0.02 | 8 |
| 11 | GPT 5 mini | OpenAI | 26.3% | $0.02 | 12 |
| 12 | Claude Haiku 4.5 | Anthropic | 22.7% | $0.61 | 100 |
| 13 | Claude Sonnet 4.6 | Anthropic | 0.0% | $44.83 | 668 |