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 | 71.2% | ok | 2113/2967 | $7.69 | 135 | 0.34h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 57.8% | ok | 1714/2967 | $4.35 | 61 | 0.23h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 57.2% | ok | 1698/2967 | $3.72 | 47 | 0.21h | 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
71.2% from 2113/2967 ProgramBench-scored tests. Raw public eval statuses: failure: 894, passed: 2287, skipped: 2.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
673c881d6dd3eval.tests.test_advanced_features.test_metamethod_le (failure)673c881d6dd3eval.tests.test_coverage_boost.test_debug_getlocal (failure)673c881d6dd3eval.tests.test_advanced_features.test_metamethod_add (failure)673c881d6dd3eval.tests.test_coverage_boost.test_string_pattern_invalid_escape (failure)673c881d6dd3eval.tests.test_coverage_boost.test_math_log_with_base (failure)673c881d6dd3eval.tests.test_coverage_boost.test_string_match_frontier (failure)
GPT 5.5 (xhigh) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260518T232006Z
57.2% from 1698/2967 ProgramBench-scored tests. Raw public eval statuses: error: 1, failure: 750, not_run: 541, passed: 1847, skipped: 2.
Agent trace: 100 shell command(s), 62 executable-observation command(s), 11 build command(s), 2 blocked attempt(s). Raw Codex logs are not published.
Likely miss class: behavioral mismatch, interactive terminal behavior mismatch.
673c881d6dd3eval.tests.test_advanced_features.test_metamethod_le (failure)673c881d6dd3eval.tests.test_coverage_boost.test_string_pattern_error (failure)673c881d6dd3eval.tests.test_coverage_boost.test_debug_getlocal (failure)673c881d6dd3eval.tests.test_advanced_features.test_metamethod_add (failure)673c881d6dd3eval.tests.test_coverage_boost.test_newproxy (failure)673c881d6dd3eval.tests.test_coverage_boost.test_collectgarbage_count (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 | 78.0% | $8.42 | 68 |
| 2 | Claude Sonnet 4.6 | Anthropic | 71.5% | $31.45 | 520 |
| 3 | GPT 5.5 (high) | OpenAI | 65.5% | $2.54 | 26 |
| 4 | Claude Opus 4.7 (xhigh) | Anthropic | 58.4% | $8.01 | 142 |
| 5 | GPT 5.5 | OpenAI | 33.9% | $1.10 | 15 |
| 6 | Claude Opus 4.6 | Anthropic | 29.0% | $19.88 | 259 |
| 7 | Gemini 3 Flash | 19.4% | $0.15 | 68 | |
| 8 | GPT 5 mini | OpenAI | 14.6% | $0.01 | 8 |
| 9 | GPT 5.4 | OpenAI | 7.7% | $0.18 | 16 |
| 10 | GPT 5.4 mini | OpenAI | 5.3% | $0.02 | 9 |
| 11 | Gemini 3.1 Pro | 3.6% | $1.22 | 122 | |
| 12 | Claude Opus 4.7 | Anthropic | 2.9% | $0.33 | 18 |
| 13 | Claude Haiku 4.5 | Anthropic | 1.2% | $0.98 | 157 |