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 | 87.0% | ok | 610/701 | $4.58 | 83 | 0.21h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 85.0% | ok | 596/701 | $3.33 | 52 | 0.18h | manifest · eval json · eval summary · agent summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 83.7% | ok | 587/701 | $3.24 | 55 | 0.21h | 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) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260518T232006Z
85.0% from 596/701 ProgramBench-scored tests. Raw public eval statuses: failure: 113, passed: 628.
Agent trace: 98 shell command(s), 74 executable-observation command(s), 6 build command(s), 2 blocked attempt(s). Raw Codex logs are not published.
Likely miss class: behavioral mismatch, exact version/output mismatch.
dbd9fc98f71ceval.tests.test_basic.test_version_command (failure)dbd9fc98f71ceval.tests.test_httpie_syntax.test_method_head (failure)dbd9fc98f71ceval.tests.test_httpie_syntax.test_form_mode_with_flag (failure)dbd9fc98f71ceval.tests.test_httpie_syntax.test_form_mode_with_long_flag (failure)15c07bf6acbdeval.tests.test_basic.test_version_command (failure)15c07bf6acbdeval.tests.test_curl_passthrough.test_compound_short_flags (failure)
manifest · eval json · eval summary · agent summary · usage audit
GPT 5.5 (xhigh) · 20260517T094700Z
83.7% from 587/701 ProgramBench-scored tests. Raw public eval statuses: failure: 121, passed: 620.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
dbd9fc98f71ceval.tests.test_basic.test_version_command (failure)15c07bf6acbdeval.tests.test_basic.test_version_command (failure)15c07bf6acbdeval.tests.test_httpie_syntax.test_url_normalization_localhost (failure)e9e6a7ad1507tests.test_binary_filter.test_late_null_byte_suppresses_entire_write (failure)e9e6a7ad1507tests.test_binary_filter.test_binary_content_null_byte_suppressed (failure)e9e6a7ad1507tests.test_binary_filter.test_multiple_writes_after_binary_ignored (failure)
Official ProgramBench Results by Model
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | GPT 5.5 (xhigh) | OpenAI | 92.9% | $7.88 | 81 |
| 2 | Claude Opus 4.7 (xhigh) | Anthropic | 92.7% | $18.14 | 157 |
| 3 | Claude Opus 4.6 | Anthropic | 89.3% | $12.49 | 243 |
| 4 | Claude Sonnet 4.6 | Anthropic | 88.4% | $25.25 | 518 |
| 5 | GPT 5.5 (high) | OpenAI | 87.0% | $3.25 | 38 |
| 6 | Claude Opus 4.7 | Anthropic | 84.9% | $15.70 | 247 |
| 7 | GPT 5.5 | OpenAI | 82.3% | $1.21 | 23 |
| 8 | Gemini 3.1 Pro | 79.0% | $1.62 | 127 | |
| 9 | Gemini 3 Flash | 70.5% | $0.22 | 85 | |
| 10 | GPT 5.4 | OpenAI | 53.5% | $0.27 | 10 |
| 11 | Claude Haiku 4.5 | Anthropic | 42.9% | $0.59 | 118 |
| 12 | GPT 5.4 mini | OpenAI | 11.1% | $0.02 | 9 |
| 13 | GPT 5 mini | OpenAI | 0.0% | $0.03 | 16 |