693generated behavioral tests
98.1%official best score
96.1%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 | 96.1% | ok | 666/693 | $2.81 | 49 | 0.15h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 10.0% | ok | 69/693 | n/a | 72 | 0.22h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 9.7% | ok | 67/693 | n/a | 57 | 0.15h | 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
96.1% from 666/693 ProgramBench-scored tests. Raw public eval statuses: failure: 27, passed: 1173.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, terminal rendering mismatch.
08291a021ea9eval.tests.test_coverage_boost.test_empty_filter_returns_error (failure)08291a021ea9eval.tests.test_config.test_htoprc_write_on_exit (failure)08291a021ea9eval.tests.test_extreme_stress.test_extreme_100_iterations (failure)08291a021ea9eval.tests.test_config_scenarios.test_run_creates_htoprc (failure)08291a021ea9eval.tests.test_extreme_stress.test_extreme_tree_100_iterations (failure)08291a021ea9eval.tests.test_config_scenarios.test_config_preserved_across_runs (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 | 98.1% | $4.18 | 47 |
| 2 | Claude Opus 4.6 | Anthropic | 85.1% | $14.15 | 238 |
| 3 | GPT 5.5 (high) | OpenAI | 84.7% | $2.63 | 28 |
| 4 | Gemini 3.1 Pro | 80.2% | $0.93 | 55 | |
| 5 | GPT 5.5 | OpenAI | 75.6% | $0.80 | 13 |
| 6 | GPT 5.4 | OpenAI | 15.3% | $0.28 | 8 |
| 7 | Claude Opus 4.7 (xhigh) | Anthropic | 9.8% | $3.13 | 104 |
| 8 | Claude Opus 4.7 | Anthropic | 9.8% | $1.55 | 59 |
| 9 | GPT 5.4 mini | OpenAI | 8.9% | $0.03 | 10 |
| 10 | Claude Sonnet 4.6 | Anthropic | 7.2% | $11.57 | 259 |
| 11 | Claude Haiku 4.5 | Anthropic | 7.2% | $0.56 | 107 |
| 12 | GPT 5 mini | OpenAI | 6.6% | $0.02 | 13 |
| 13 | Gemini 3 Flash | 6.5% | $0.21 | 43 |