502generated behavioral tests
88.2%official best score
88.4%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 | 88.4% | ok | 444/502 | $3.30 | 67 | 0.18h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 85.7% | ok | 430/502 | n/a | 54 | 0.16h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 85.1% | ok | 427/502 | n/a | 37 | 0.12h | 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
88.4% from 444/502 ProgramBench-scored tests. Raw public eval statuses: failure: 217, passed: 760, skipped: 2.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch, terminal rendering mismatch.
1b6c3c2e2d37eval.tests.test_tui.TestTUICoverageFinal.test_tui_register_show_command (failure)30cdbea7de0atests.test_cli_additional.test_config_malformed_yaml_graceful_fallback (failure)30cdbea7de0atests.test_errors_extended.test_yaml_error_invalid_color_value (failure)30cdbea7de0atests.test_errors_extended.test_yaml_error_invalid_config (failure)30cdbea7de0atests.test_cli_additional.test_config_wrong_types_graceful_fallback (failure)30cdbea7de0atests.test_errors_extended.test_multiple_yaml_errors (failure)
GPT 5.5 (xhigh) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
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 | GPT 5.5 (high) | OpenAI | 88.2% | $3.14 | 51 |
| 2 | GPT 5.5 | OpenAI | 88.0% | $1.12 | 22 |
| 3 | Claude Opus 4.7 (xhigh) | Anthropic | 76.5% | $1.68 | 52 |
| 4 | GPT 5.5 (xhigh) | OpenAI | 56.8% | $6.12 | 46 |
| 5 | Gemini 3.1 Pro | 49.2% | $1.12 | 62 | |
| 6 | GPT 5.4 mini | OpenAI | 46.0% | $0.06 | 44 |
| 7 | Claude Opus 4.7 | Anthropic | 45.2% | $1.13 | 52 |
| 8 | GPT 5.4 | OpenAI | 39.0% | $0.42 | 12 |
| 9 | Gemini 3 Flash | 33.3% | $0.27 | 76 | |
| 10 | Claude Haiku 4.5 | Anthropic | 23.5% | $1.01 | 142 |
| 11 | GPT 5 mini | OpenAI | 22.3% | $0.05 | 20 |
| 12 | Claude Opus 4.6 | Anthropic | 0.0% | $2.48 | 244 |
| 13 | Claude Sonnet 4.6 | Anthropic | 0.0% | $4.16 | 360 |