531generated behavioral tests
95.7%official best score
81.0%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) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 81.0% | ok | 430/531 | $5.72 | 82 | 0.29h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 79.3% | ok | 421/531 | n/a | 77 | 0.30h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 78.9% | ok | 419/531 | 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
81.0% from 430/531 ProgramBench-scored tests. Raw public eval statuses: failure: 105, passed: 524, skipped: 1.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
f959760f19f7eval.tests.test_edge_cases.test_total_width_very_small (failure)f959760f19f7eval.tests.test_measurement_options.test_quantity_short_flag (failure)231dfd9dcc35eval.tests.test_advanced_errors.test_column_width_zero_values (failure)231dfd9dcc35eval.tests.test_advanced_errors.test_total_width_very_small (failure)dc10e832e5bbtests.test_flag_interactions.test_total_width_constrains_output (failure)dc10e832e5bbtests.test_gap_fill.test_overlapping_parent_child_keeps_parent (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 (xhigh) | OpenAI | 95.7% | $6.67 | 57 |
| 2 | Claude Opus 4.6 | Anthropic | 86.1% | $18.23 | 273 |
| 3 | Claude Sonnet 4.6 | Anthropic | 85.1% | $29.37 | 456 |
| 4 | Claude Opus 4.7 (xhigh) | Anthropic | 83.4% | $13.81 | 153 |
| 5 | GPT 5.5 (high) | OpenAI | 82.7% | $3.38 | 32 |
| 6 | Claude Opus 4.7 | Anthropic | 76.5% | $4.66 | 99 |
| 7 | GPT 5.5 | OpenAI | 73.6% | $1.01 | 14 |
| 8 | GPT 5.4 | OpenAI | 63.3% | $0.60 | 14 |
| 9 | Gemini 3 Flash | 57.3% | $0.42 | 95 | |
| 10 | Claude Haiku 4.5 | Anthropic | 55.4% | $1.02 | 140 |
| 11 | Gemini 3.1 Pro | 54.4% | $2.34 | 129 | |
| 12 | GPT 5.4 mini | OpenAI | 52.9% | $0.03 | 11 |
| 13 | GPT 5 mini | OpenAI | 13.7% | $0.02 | 13 |