ProgramBench task detail

mgdm__htmlq.6e31bc8

Task-level results for this Codex /goal scaffold. ProgramBench baseline context is cached from the official task page; Codex scored tests are after active-branch and ignored-test filtering.

1455generated behavioral tests
98.0%official best score
95.1%Codex best score
3Codex result rows

Codex Results by Model

#ModelRunAgentModeScoreEvalTestsEst. costCallsWallEvidence
1 GPT 5.5 (xhigh) 20260517T094700Z Codex /goal Mini-SWE-compatible no internet 95.1% ok 1383/1455 $5.11 93 0.24h manifest · eval json · eval summary · usage audit
2 GPT 5.5 (xhigh) 20260518T232006Z Codex /goal Paper prompt + /goal no internet 94.4% ok 1374/1455 $3.37 61 0.20h manifest · eval json · eval summary · agent summary · usage audit
3 GPT 5.5 (xhigh) 20260519T215046Z Codex /goal Paper prompt + Goal contract no internet 93.7% ok 1364/1455 $4.47 78 0.25h 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

95.1% from 1383/1455 ProgramBench-scored tests. Raw public eval statuses: failure: 72, passed: 1984, skipped: 2.

Agent trace summary unavailable.

Likely miss class: behavioral mismatch.

  • 61b7ba1e3b42 eval.tests.test_09_urls_and_links.test_link_with_base (failure)
  • 61b7ba1e3b42 eval.tests.test_07_massive_parametrized.test_base_url_rewriting[https://example.com-/path-https://example.com/path] (failure)
  • 61b7ba1e3b42 eval.tests.test_04_flags.test_base_flag (failure)
  • 61b7ba1e3b42 eval.tests.test_09_urls_and_links.test_base_tag_detection (failure)
  • 61b7ba1e3b42 eval.tests.test_07_massive_parametrized.test_base_url_rewriting[https://example.com/-/path-https://example.com/path] (failure)
  • 61b7ba1e3b42 eval.tests.test_04_flags.test_base_flag_short (failure)

manifest · eval json · eval summary · usage audit

GPT 5.5 (xhigh) · 20260518T232006Z

94.4% from 1374/1455 ProgramBench-scored tests. Raw public eval statuses: failure: 81, passed: 1975, skipped: 2.

Agent trace: 178 shell command(s), 154 executable-observation command(s), 7 build command(s), 2 blocked attempt(s). Raw Codex logs are not published.

Likely miss class: behavioral mismatch.

  • b919df31faac eval.tests.test_remove_nodes.test_remove_nodes_all_filtered_out (failure)
  • b919df31faac eval.tests.test_remove_nodes.test_remove_nodes_filters_selected_elements (failure)
  • b9ea5e667678 tests.test_cli_edge.test_multiple_attribute_flags_error (failure)
  • b9ea5e667678 tests.test_gaps.test_empty_href_attribute (failure)
  • b9ea5e667678 tests.test_cli_edge.test_pretty_print_formatting (failure)
  • b9ea5e667678 tests.test_gaps.test_pretty_print_with_comments (failure)

manifest · eval json · eval summary · agent summary · usage audit

GPT 5.5 (xhigh) · 20260519T215046Z

Public eval evidence has not been exported for this row yet.

Official ProgramBench Results by Model

#ModelProviderScoreCostCalls
1 GPT 5.5 (xhigh) OpenAI 98.0% $5.09 69
2 GPT 5.5 (high) OpenAI 96.4% $3.62 52
3 Claude Opus 4.7 (xhigh) Anthropic 96.2% $15.99 192
4 Claude Opus 4.6 Anthropic 93.9% $13.55 325
5 Claude Opus 4.7 Anthropic 92.3% $3.11 108
6 Claude Sonnet 4.6 Anthropic 88.0% $22.96 627
7 Gemini 3.1 Pro Google 87.6% $1.46 104
8 Gemini 3 Flash Google 84.4% $0.29 140
9 GPT 5.5 OpenAI 84.3% $0.99 17
10 Claude Haiku 4.5 Anthropic 78.0% $0.58 144
11 GPT 5.4 OpenAI 60.8% $0.30 14
12 GPT 5.4 mini OpenAI 37.8% $0.03 9
13 GPT 5 mini OpenAI 1.2% $0.02 13