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 | 87.3% | ok | 773/885 | $3.76 | 62 | 0.22h | manifest · eval json · eval summary · agent summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 23.5% | ok | 208/885 | $4.09 | 57 | 0.23h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 3.8% | ok | 34/885 | $4.54 | 90 | 0.23h | manifest · eval json · eval 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) · 20260518T232006Z
87.3% from 773/885 ProgramBench-scored tests. Raw public eval statuses: failure: 118, passed: 856.
Agent trace: 133 shell command(s), 115 executable-observation command(s), 3 build command(s), 2 blocked attempt(s). Raw Codex logs are not published.
Likely miss class: behavioral mismatch, exact version/output mismatch.
9e91d4ba23b1eval.tests.test_domain.test_domain_without_protocol (failure)9e91d4ba23b1eval.tests.test_basic_invocation.test_stdin_empty_input (failure)9e91d4ba23b1eval.tests.test_input_output.test_output_file_no_overwrite (failure)9e91d4ba23b1eval.tests.test_input_output.test_output_file_with_overwrite (failure)9e91d4ba23b1eval.tests.test_input_output.test_output_to_file (failure)9e91d4ba23b1eval.tests.test_input_output.test_empty_input_flag_value (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.
GPT 5.5 (xhigh) · 20260517T094700Z
3.8% from 34/885 ProgramBench-scored tests. Raw public eval statuses: failure: 926, passed: 48.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
9e91d4ba23b1eval.tests.test_advanced_features.test_blockquote_nested (failure)9e91d4ba23b1eval.tests.test_advanced_features.test_preformatted_text (failure)9e91d4ba23b1eval.tests.test_advanced_features.test_url_with_query_params (failure)9e91d4ba23b1eval.tests.test_advanced_features.test_nested_lists (failure)9e91d4ba23b1eval.tests.test_advanced_features.test_cite_tag (failure)9e91d4ba23b1eval.tests.test_advanced_features.test_relative_url_with_dots (failure)
Official ProgramBench Results by Model
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | GPT 5.5 (xhigh) | OpenAI | 92.8% | $7.12 | 72 |
| 2 | GPT 5.5 (high) | OpenAI | 88.2% | $2.38 | 29 |
| 3 | Claude Opus 4.7 | Anthropic | 85.5% | $3.56 | 92 |
| 4 | Claude Opus 4.6 | Anthropic | 83.7% | $14.48 | 363 |
| 5 | Claude Sonnet 4.6 | Anthropic | 81.9% | $18.69 | 503 |
| 6 | Claude Opus 4.7 (xhigh) | Anthropic | 81.1% | $10.37 | 209 |
| 7 | GPT 5.5 | OpenAI | 77.7% | $0.78 | 15 |
| 8 | GPT 5.4 | OpenAI | 67.2% | $0.23 | 7 |
| 9 | Gemini 3 Flash | 65.9% | $0.39 | 105 | |
| 10 | Claude Haiku 4.5 | Anthropic | 47.3% | $0.72 | 111 |
| 11 | GPT 5.4 mini | OpenAI | 33.2% | $0.05 | 11 |
| 12 | Gemini 3.1 Pro | 1.5% | $1.38 | 113 | |
| 13 | GPT 5 mini | OpenAI | 1.5% | $0.06 | 31 |