1224generated behavioral tests
76.7%official best score
69.0%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) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 69.0% | ok | 845/1224 | n/a | 46 | 0.16h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 55.3% | ok | 677/1224 | $4.35 | 63 | 0.21h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 32.7% | ok | 400/1224 | n/a | 83 | 0.31h | 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) · 20260518T232006Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260517T094700Z
55.3% from 677/1224 ProgramBench-scored tests. Raw public eval statuses: failure: 698, passed: 1014, skipped: 3.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, interactive terminal behavior mismatch, terminal rendering mismatch.
eb89f66d8b70eval.tests.test_basic.test_empty_input_with_exit_0_flag (failure)eb89f66d8b70eval.tests.test_basic.test_initial_filter_regexp_no_match (failure)eb89f66d8b70eval.tests.test_basic.test_multiple_terms_no_match (failure)eb89f66d8b70eval.tests.test_basic.test_initial_filter_case_sensitive_no_match (failure)eb89f66d8b70eval.tests.test_basic.test_query_no_match_single_line (failure)eb89f66d8b70eval.tests.test_basic.test_regexp_filter_no_match_due_to_anchor (failure)
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 4.6 | Anthropic | 76.7% | $14.19 | 391 |
| 2 | Claude Sonnet 4.6 | Anthropic | 74.6% | $15.96 | 351 |
| 3 | Claude Opus 4.7 (xhigh) | Anthropic | 72.3% | $9.13 | 157 |
| 4 | Gemini 3.1 Pro | 71.4% | $1.57 | 74 | |
| 5 | GPT 5.5 (high) | OpenAI | 69.4% | $4.35 | 46 |
| 6 | GPT 5.5 (xhigh) | OpenAI | 68.6% | $8.02 | 84 |
| 7 | Claude Opus 4.7 | Anthropic | 67.5% | $2.50 | 84 |
| 8 | GPT 5.4 | OpenAI | 66.5% | $0.29 | 11 |
| 9 | GPT 5.5 | OpenAI | 65.0% | $1.13 | 16 |
| 10 | GPT 5 mini | OpenAI | 56.6% | $0.03 | 10 |
| 11 | GPT 5.4 mini | OpenAI | 28.9% | $0.03 | 8 |
| 12 | Claude Haiku 4.5 | Anthropic | 16.3% | $1.14 | 146 |
| 13 | Gemini 3 Flash | 11.6% | $0.20 | 45 |