281generated behavioral tests
86.5%official best score
97.2%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) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 97.2% | ok | 273/281 | $4.11 | 61 | 0.20h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 58.0% | ok | 163/281 | n/a | 44 | 0.17h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 56.6% | ok | 159/281 | n/a | 49 | 0.17h | 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
97.2% from 273/281 ProgramBench-scored tests. Raw public eval statuses: failure: 17, passed: 300, skipped: 2.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, interactive terminal behavior mismatch, terminal rendering mismatch.
ed5c2b1ffc48tests.test_edge_cases.test_custom_date_format_changes_display (failure)ed5c2b1ffc48tests.test_edge_cases.test_screensaver_mode_quits_on_keypress (failure)ed5c2b1ffc48tests.test_edge_cases.test_movement_blocked_when_centered (failure)ed5c2b1ffc48tests.test_gap_fill.test_screensaver_color_change_key_1 (failure)ed5c2b1ffc48tests.test_gap_fill.test_screensaver_color_change_key_5 (failure)ed5c2b1ffc48tests.test_tui_interactive.test_movement_arrow_keys_down (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 | Claude Opus 4.7 (xhigh) | Anthropic | 86.5% | $10.01 | 138 |
| 2 | Claude Opus 4.6 | Anthropic | 84.0% | $10.30 | 218 |
| 3 | GPT 5.5 (xhigh) | OpenAI | 83.6% | $7.83 | 74 |
| 4 | GPT 5.5 (high) | OpenAI | 82.9% | $4.43 | 50 |
| 5 | GPT 5.5 | OpenAI | 77.6% | $1.24 | 16 |
| 6 | GPT 5.4 | OpenAI | 76.9% | $0.22 | 8 |
| 7 | Claude Opus 4.7 | Anthropic | 69.4% | $5.70 | 123 |
| 8 | Claude Sonnet 4.6 | Anthropic | 64.4% | $12.01 | 192 |
| 9 | Gemini 3.1 Pro | 63.0% | $2.35 | 90 | |
| 10 | Gemini 3 Flash | 58.4% | $0.28 | 45 | |
| 11 | GPT 5.4 mini | OpenAI | 39.9% | $0.02 | 6 |
| 12 | GPT 5 mini | OpenAI | 21.7% | $0.01 | 10 |
| 13 | Claude Haiku 4.5 | Anthropic | 5.3% | $0.88 | 139 |