ProgramBench task detail

eudoxia0__hashcards.48aa136

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.

1019generated behavioral tests
86.8%official best score
82.9%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 82.9% ok 845/1019 $7.12 128 0.30h manifest · eval json · eval summary · usage audit
2 GPT 5.5 (xhigh) 20260518T232006Z Codex /goal Paper prompt + /goal no internet 78.2% ok 797/1019 n/a 107 0.27h manifest · eval json · eval summary · agent summary · usage audit
3 GPT 5.5 (xhigh) 20260519T215046Z Codex /goal Paper prompt + Goal contract no internet 70.7% ok 720/1019 n/a 106 0.28h 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

82.9% from 845/1019 ProgramBench-scored tests. Raw public eval statuses: failure: 217, passed: 1072, skipped: 4.

Agent trace summary unavailable.

Likely miss class: behavioral mismatch.

  • 8c389ff5af21 tests.test_card_parsing.test_frontmatter_custom_deck_name (failure)
  • 8c389ff5af21 tests.test_error_handling.test_unclosed_frontmatter (failure)
  • 8c389ff5af21 tests.test_error_handling.test_malformed_frontmatter (failure)
  • 8c389ff5af21 tests.test_error_handling.test_stats_with_broken_symlink (failure)
  • 8c389ff5af21 tests.test_export_command.test_export_with_frontmatter (failure)
  • 8c389ff5af21 tests.test_media_handling.test_image_with_title (failure)

manifest · eval json · eval summary · usage audit

GPT 5.5 (xhigh) · 20260518T232006Z

78.2% from 797/1019 ProgramBench-scored tests. Raw public eval statuses: failure: 292, passed: 997, skipped: 4.

Agent trace: 162 shell command(s), 94 executable-observation command(s), 11 build command(s), 1 blocked attempt(s). Raw Codex logs are not published.

Likely miss class: behavioral mismatch.

  • 8c389ff5af21 tests.test_card_parsing.test_frontmatter_custom_deck_name (failure)
  • 8c389ff5af21 tests.test_error_handling.test_unclosed_frontmatter (failure)
  • 8c389ff5af21 tests.test_error_handling.test_malformed_frontmatter (failure)
  • 8c389ff5af21 tests.test_database_operations.test_duplicate_cards_deduplicated (failure)
  • 8c389ff5af21 tests.test_error_handling.test_stats_with_broken_symlink (failure)
  • 8c389ff5af21 tests.test_export_command.test_export_with_frontmatter (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 (high) OpenAI 86.8% $5.14 63
2 GPT 5.5 (xhigh) OpenAI 82.5% $9.67 100
3 GPT 5.5 OpenAI 77.6% $2.18 28
4 Gemini 3 Flash Google 56.3% $0.33 114
5 Claude Haiku 4.5 Anthropic 51.5% $0.82 141
6 GPT 5.4 mini OpenAI 41.5% $0.04 7
7 GPT 5 mini OpenAI 38.0% $0.04 25
8 Claude Opus 4.7 Anthropic 6.1% $1.55 65
9 Claude Sonnet 4.6 Anthropic 6.1% $51.19 537
10 Claude Opus 4.7 (xhigh) Anthropic 3.7% $11.74 179
11 GPT 5.4 OpenAI 3.7% $0.48 15
12 Claude Opus 4.6 Anthropic 0.0% $6.91 254
13 Gemini 3.1 Pro Google 0.0% $2.07 136