470generated behavioral tests
89.6%official best score
75.1%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) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 75.1% | ok | 353/470 | n/a | 52 | 0.15h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 55.3% | ok | 260/470 | $3.34 | 56 | 0.20h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 36.0% | ok | 169/470 | n/a | 40 | 0.11h | 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
55.3% from 260/470 ProgramBench-scored tests. Raw public eval statuses: failure: 258, passed: 527, skipped: 1.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
a56ac371fb39eval.tests.test_basic_invocation.TestPathHandling.test_absolute_path_argument (failure)a56ac371fb39eval.tests.test_basic_invocation.TestFlags.test_fuzzy_flag (failure)a56ac371fb39eval.tests.test_basic_invocation.TestFlags.test_dir_only_flag (failure)a56ac371fb39eval.tests.test_basic_invocation.TestFlags.test_all_flags_combined (failure)a56ac371fb39eval.tests.test_basic_invocation.TestHelpAndVersion.test_help_flag_long (failure)a56ac371fb39eval.tests.test_edge_cases.TestFlagCombinations.test_icons_and_preview (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 | GPT 5.5 (xhigh) | OpenAI | 89.6% | $7.96 | 55 |
| 2 | GPT 5.5 (high) | OpenAI | 83.8% | $4.11 | 45 |
| 3 | GPT 5.5 | OpenAI | 80.4% | $1.30 | 20 |
| 4 | Claude Opus 4.7 | Anthropic | 74.3% | $1.65 | 70 |
| 5 | Claude Opus 4.7 (xhigh) | Anthropic | 72.3% | $5.10 | 118 |
| 6 | Claude Opus 4.6 | Anthropic | 58.9% | $15.76 | 297 |
| 7 | Claude Sonnet 4.6 | Anthropic | 48.3% | $23.52 | 407 |
| 8 | Claude Haiku 4.5 | Anthropic | 44.0% | $0.88 | 175 |
| 9 | GPT 5.4 | OpenAI | 30.2% | $0.24 | 13 |
| 10 | Gemini 3 Flash | 26.2% | $0.16 | 35 | |
| 11 | GPT 5.4 mini | OpenAI | 26.0% | $0.02 | 7 |
| 12 | Gemini 3.1 Pro | 22.6% | $1.63 | 80 | |
| 13 | GPT 5 mini | OpenAI | 22.3% | $0.02 | 16 |