978generated behavioral tests
87.8%official best score
86.7%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 | 86.7% | ok | 848/978 | $3.26 | 49 | 0.19h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 84.6% | ok | 827/978 | n/a | 44 | 0.16h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 36.9% | ok | 361/978 | n/a | 39 | 0.18h | 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
86.7% from 848/978 ProgramBench-scored tests. Raw public eval statuses: failure: 132, passed: 989, skipped: 6.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
d14b27faa8deeval.tests.test_advanced_substitution.test_substitution_with_null_value (failure)d14b27faa8deeval.tests.test_additional_coverage.test_print_lua_flag (failure)d14b27faa8deeval.tests.test_completions.test_generate_completions_bash (failure)d14b27faa8deeval.tests.test_completions.test_generate_completions_zsh (failure)d14b27faa8deeval.tests.test_debugging.test_print_lua_with_filter (failure)d14b27faa8deeval.tests.test_debugging.test_print_lua (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 (high) | OpenAI | 87.8% | $2.71 | 39 |
| 2 | GPT 5.5 (xhigh) | OpenAI | 86.2% | $5.98 | 53 |
| 3 | Claude Opus 4.6 | Anthropic | 86.0% | $12.29 | 304 |
| 4 | Claude Opus 4.7 | Anthropic | 74.4% | $5.12 | 127 |
| 5 | GPT 5.5 | OpenAI | 70.0% | $0.83 | 12 |
| 6 | Gemini 3.1 Pro | 66.7% | $1.92 | 122 | |
| 7 | Claude Haiku 4.5 | Anthropic | 38.9% | $0.96 | 144 |
| 8 | GPT 5 mini | OpenAI | 28.9% | $0.02 | 20 |
| 9 | GPT 5.4 mini | OpenAI | 10.1% | $0.05 | 20 |
| 10 | Claude Opus 4.7 (xhigh) | Anthropic | 1.3% | $6.80 | 125 |
| 11 | Claude Sonnet 4.6 | Anthropic | 1.3% | $13.91 | 400 |
| 12 | GPT 5.4 | OpenAI | 1.2% | $0.24 | 8 |
| 13 | Gemini 3 Flash | 1.2% | $0.24 | 77 |