1724generated behavioral tests
32.7%official best score
41.5%Codex best score
3Codex result rows
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 | 41.5% | ok | 716/1724 | $3.90 | 54 | 0.20h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 40.7% | ok | 701/1724 | n/a | 45 | 0.17h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 39.3% | ok | 677/1724 | n/a | 51 | 0.23h | 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
41.5% from 716/1724 ProgramBench-scored tests. Raw public eval statuses: failure: 1025, passed: 760, skipped: 1.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
2f270486dd38eval.tests.test_additional_coverage.test_compile_with_figures (failure)2f270486dd38eval.tests.test_additional_coverage.test_eval_boolean_operations (failure)2f270486dd38eval.tests.test_advanced_compile.test_compile_multiple_inputs (failure)2f270486dd38eval.tests.test_advanced_compile.test_compile_with_typst_code (failure)2f270486dd38eval.tests.test_compile.test_compile_from_stdin (failure)2f270486dd38eval.tests.test_compile.test_compile_to_stdout (failure)
GPT 5.5 (xhigh) · 20260518T232006Z
Public eval evidence has not been exported for this row yet.
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 | OpenAI | 32.7% | $0.75 | 13 |
| 2 | GPT 5.4 | OpenAI | 28.0% | $0.39 | 11 |
| 3 | Claude Haiku 4.5 | Anthropic | 20.6% | $0.72 | 127 |
| 4 | Gemini 3 Flash | 15.8% | $0.28 | 75 | |
| 5 | Gemini 3.1 Pro | 15.5% | $0.67 | 44 | |
| 6 | Claude Opus 4.7 (xhigh) | Anthropic | 14.4% | $5.44 | 125 |
| 7 | Claude Opus 4.6 | Anthropic | 13.3% | $4.62 | 285 |
| 8 | Claude Opus 4.7 | Anthropic | 11.7% | $1.51 | 50 |
| 9 | GPT 5 mini | OpenAI | 3.8% | $0.02 | 8 |
| 10 | GPT 5.5 (xhigh) | OpenAI | 3.7% | $10.06 | 83 |
| 11 | GPT 5.4 mini | OpenAI | 2.7% | $0.03 | 7 |
| 12 | GPT 5.5 (high) | OpenAI | 1.3% | $3.71 | 32 |
| 13 | Claude Sonnet 4.6 | Anthropic | 0.0% | $52.50 | 855 |