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 | 52.9% | ok | 1495/2828 | $4.59 | 80 | 0.27h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 21.4% | ok | 604/2828 | $3.65 | 57 | 0.20h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 12.0% | ok | 339/2828 | $3.96 | 74 | 0.22h | manifest · eval json · eval summary · agent summary · usage audit |
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
52.9% from 1495/2828 ProgramBench-scored tests. Raw public eval statuses: failure: 1354, passed: 1796, skipped: 5.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
65648d8896daeval.tests.test_advanced_features.test_page_rules (failure)65648d8896daeval.tests.test_browserslist.test_browserslistrc_discovery (failure)65648d8896daeval.tests.test_cli_combinations.test_minify_with_sourcemap (failure)65648d8896daeval.tests.test_css_at_rules_comprehensive.test_page_at_rule (failure)65648d8896daeval.tests.test_comprehensive_css.test_vendor_prefix_removal (failure)65648d8896daeval.tests.test_css_modern_features.test_print_media_comprehensive (failure)
GPT 5.5 (xhigh) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260518T232006Z
12.0% from 339/2828 ProgramBench-scored tests. Raw public eval statuses: failure: 2753, passed: 397, skipped: 5.
Agent trace: 140 shell command(s), 94 executable-observation command(s), 6 build command(s), 1 blocked attempt(s). Raw Codex logs are not published.
Likely miss class: behavioral mismatch.
65648d8896daeval.tests.test_advanced_features.test_page_rules (failure)65648d8896daeval.tests.test_advanced_features.test_counter_style (failure)65648d8896daeval.tests.test_css_properties_massive.test_property_value[animation-fill-mode-forwards] (failure)65648d8896daeval.tests.test_css_properties_massive.test_property_value[border-style-ridge] (failure)65648d8896daeval.tests.test_css_properties_comprehensive.test_custom_properties (failure)65648d8896daeval.tests.test_comprehensive_css.test_minify_shorthand_combination (failure)
manifest · eval json · eval summary · agent summary · usage audit
Official ProgramBench Results by Model
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | GPT 5.5 (xhigh) | OpenAI | 55.0% | $8.05 | 68 |
| 2 | Claude Opus 4.6 | Anthropic | 53.6% | $17.40 | 312 |
| 3 | GPT 5.5 (high) | OpenAI | 52.7% | $3.53 | 40 |
| 4 | Claude Sonnet 4.6 | Anthropic | 49.9% | $45.82 | 705 |
| 5 | Claude Opus 4.7 (xhigh) | Anthropic | 46.5% | $4.45 | 107 |
| 6 | GPT 5.5 | OpenAI | 44.9% | $1.23 | 20 |
| 7 | Claude Opus 4.7 | Anthropic | 43.2% | $1.26 | 46 |
| 8 | Gemini 3 Flash | 43.0% | $0.33 | 117 | |
| 9 | GPT 5.4 | OpenAI | 36.7% | $0.25 | 12 |
| 10 | GPT 5 mini | OpenAI | 36.6% | $0.03 | 23 |
| 11 | Gemini 3.1 Pro | 34.0% | $1.22 | 91 | |
| 12 | GPT 5.4 mini | OpenAI | 31.8% | $0.03 | 22 |
| 13 | Claude Haiku 4.5 | Anthropic | 28.7% | $1.29 | 165 |