Codex Results by Model
| # | Model | Run | Agent | Mode | Score | Eval | Tests | Est. cost | Calls | Wall | Evidence |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 50.3% | ok | 340/676 | $4.87 | 62 | 0.23h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 45.1% | ok | 305/676 | $9.83 | 113 | 0.40h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 0.0% | error | 0/676 | $5.20 | 80 | 0.24h | 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) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260517T094700Z
45.1% from 305/676 ProgramBench-scored tests. Raw public eval statuses: failure: 365, passed: 424, skipped: 50.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, interactive terminal behavior mismatch.
df88744325cfeval.tests.test_age_advanced.TestDuplicates.test_duplicate_identity_file_warns (failure)df88744325cfeval.tests.test_age_advanced.TestDuplicates.test_duplicate_recipient_warns (failure)df88744325cfeval.tests.test_age_batchpass.TestBatchpassBasic.test_batchpass_encrypt_decrypt (failure)df88744325cfeval.tests.test_age_coverage_boost.TestRSAFullRoundtrip.test_encrypt_to_rsa_decrypt_with_identity (failure)df88744325cfeval.tests.test_age_batchpass.TestBatchpassBasic.test_batchpass_decrypt_with_d_flag (failure)df88744325cfeval.tests.test_age_coverage_boost.TestPostQuantumRoundtrip.test_encrypt_decrypt_pq_roundtrip (failure)
GPT 5.5 (xhigh) · 20260518T232006Z
0.0% from 0/676 ProgramBench-scored tests. Raw public eval statuses: not_run: 839.
Agent trace: 102 shell command(s), 47 executable-observation command(s), 14 build command(s), 0 blocked attempt(s). Raw Codex logs are not published.
Likely miss class: behavioral mismatch, exact version/output mismatch, interactive terminal behavior mismatch.
df88744325cfeval.tests.test_age_advanced.TestDuplicates.test_duplicate_identity_file_warns (not_run)df88744325cfeval.tests.test_age_advanced.TestDuplicates.test_duplicate_recipient_warns (not_run)df88744325cfeval.tests.test_age_advanced.TestDuplicates.test_duplicate_recipients_file_warns (not_run)df88744325cfeval.tests.test_age_advanced.TestDuplicates.test_no_duplicate_no_warning (not_run)df88744325cfeval.tests.test_age_advanced.TestEncryptToIdentity.test_encrypt_and_decrypt_with_identity (not_run)df88744325cfeval.tests.test_age_advanced.TestEncryptToIdentity.test_encrypt_to_identity_and_additional_recipient (not_run)
manifest · eval json · eval summary · agent summary · usage audit
Official ProgramBench Results by Model
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 4.6 | Anthropic | 63.5% | $43.00 | 547 |
| 2 | Claude Opus 4.6 | Anthropic | 60.9% | $13.80 | 210 |
| 3 | GPT 5.5 (high) | OpenAI | 60.5% | $6.36 | 60 |
| 4 | GPT 5.5 (xhigh) | OpenAI | 54.9% | $7.10 | 66 |
| 5 | Gemini 3.1 Pro | 49.3% | $1.41 | 100 | |
| 6 | GPT 5.5 | OpenAI | 38.5% | $1.07 | 15 |
| 7 | Claude Opus 4.7 | Anthropic | 13.9% | $1.05 | 37 |
| 8 | Claude Haiku 4.5 | Anthropic | 8.9% | $0.91 | 156 |
| 9 | GPT 5.4 | OpenAI | 8.6% | $0.47 | 14 |
| 10 | Gemini 3 Flash | 3.1% | $0.22 | 66 | |
| 11 | GPT 5.4 mini | OpenAI | 0.3% | $0.04 | 8 |
| 12 | GPT 5 mini | OpenAI | 0.1% | $0.02 | 13 |
| 13 | Claude Opus 4.7 (xhigh) | Anthropic | 0.0% | $4.35 | 65 |