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 | 90.5% | ok | 553/611 | $4.55 | 98 | 0.24h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 88.2% | ok | 539/611 | $3.67 | 74 | 0.23h | manifest · eval json · eval summary · agent summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 85.9% | ok | 525/611 | $2.80 | 55 | 0.17h | 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
90.5% from 553/611 ProgramBench-scored tests. Raw public eval statuses: failure: 65, passed: 870, skipped: 2.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
1b0359e8d221tests.test_extensive_patterns.test_alternation_in_patterns (failure)1b0359e8d221tests.test_ignore_mechanisms.test_pragma_allowlist_secret (failure)1b0359e8d221tests.test_precommit.test_install_precommit_with_precommit_d (failure)1b0359e8d221tests.test_precommit.test_install_precommit_already_installed (failure)1b0359e8d221tests.test_ignore_mechanisms.test_pragma_allowlist_secret_inline (failure)1b0359e8d221tests.test_precommit.test_install_precommit_with_strict_ignore_error (failure)
GPT 5.5 (xhigh) · 20260518T232006Z
88.2% from 539/611 ProgramBench-scored tests. Raw public eval statuses: failure: 86, passed: 849, skipped: 2.
Agent trace: 143 shell command(s), 94 executable-observation command(s), 6 build command(s), 2 blocked attempt(s). Raw Codex logs are not published.
Likely miss class: behavioral mismatch.
1b0359e8d221tests.test_basic.test_nonexistent_file (failure)1b0359e8d221tests.test_extensive_patterns.test_alternation_in_patterns (failure)1b0359e8d221tests.test_ignore_mechanisms.test_pragma_allowlist_secret (failure)1b0359e8d221tests.test_precommit.test_install_precommit_with_strict_ignore_error (failure)1b0359e8d221tests.test_precommit.test_install_precommit_appends_to_existing (failure)1b0359e8d221tests.test_ignore_mechanisms.test_pragma_allowlist_secret_inline (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
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | GPT 5.5 (high) | OpenAI | 91.2% | $4.05 | 52 |
| 2 | GPT 5.5 | OpenAI | 88.7% | $0.74 | 16 |
| 3 | GPT 5.5 (xhigh) | OpenAI | 87.7% | $6.58 | 77 |
| 4 | Claude Opus 4.7 (xhigh) | Anthropic | 77.4% | $15.13 | 210 |
| 5 | Claude Opus 4.7 | Anthropic | 72.8% | $7.44 | 156 |
| 6 | Claude Opus 4.6 | Anthropic | 63.8% | $16.13 | 268 |
| 7 | Gemini 3 Flash | 59.9% | $0.39 | 129 | |
| 8 | GPT 5.4 | OpenAI | 46.3% | $0.24 | 12 |
| 9 | GPT 5 mini | OpenAI | 42.9% | $0.01 | 16 |
| 10 | Gemini 3.1 Pro | 19.6% | $2.14 | 115 | |
| 11 | Claude Haiku 4.5 | Anthropic | 18.0% | $1.51 | 186 |
| 12 | Claude Sonnet 4.6 | Anthropic | 7.4% | $65.81 | 845 |
| 13 | GPT 5.4 mini | OpenAI | 0.0% | $0.04 | 10 |