14637generated behavioral tests
24.1%official best score
32.1%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 | 32.1% | ok | 4694/14637 | $7.10 | 97 | 0.36h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 28.3% | ok | 4147/14637 | n/a | 41 | 0.29h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 27.6% | ok | 4044/14637 | n/a | 63 | 0.27h | 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
32.1% from 4694/14637 ProgramBench-scored tests. Raw public eval statuses: failure: 10248, passed: 5779, skipped: 2.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, interactive terminal behavior mismatch.
c55c103e74aeeval.tests.test_coverage_boost_50.test_reorder_verb (failure)c55c103e74aeeval.tests.test_absolute_final.test_a36 (failure)c55c103e74aeeval.tests.test_basic_invocation.test_help_list_functions (failure)c55c103e74aeeval.tests.test_absolute_final_push_50.test_acosh_function (failure)c55c103e74aeeval.tests.test_coverage_boost_50.test_sec2gmt_verb (failure)c55c103e74aeeval.tests.test_absolute_final_push_50.test_contains_true (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 | Claude Opus 4.7 (xhigh) | Anthropic | 24.1% | $8.12 | 127 |
| 2 | GPT 5.5 (high) | OpenAI | 23.4% | $3.02 | 26 |
| 3 | Claude Opus 4.6 | Anthropic | 22.9% | $13.86 | 322 |
| 4 | Claude Opus 4.7 | Anthropic | 19.0% | $2.86 | 62 |
| 5 | GPT 5.5 | OpenAI | 16.9% | $1.02 | 15 |
| 6 | Gemini 3 Flash | 11.5% | $0.20 | 58 | |
| 7 | GPT 5.4 | OpenAI | 8.0% | $0.32 | 11 |
| 8 | Claude Haiku 4.5 | Anthropic | 4.9% | $1.61 | 206 |
| 9 | Gemini 3.1 Pro | 4.8% | $0.95 | 59 | |
| 10 | GPT 5.4 mini | OpenAI | 4.7% | $0.02 | 8 |
| 11 | GPT 5 mini | OpenAI | 3.4% | $0.02 | 12 |
| 12 | GPT 5.5 (xhigh) | OpenAI | 0.0% | $5.15 | 60 |
| 13 | Claude Sonnet 4.6 | Anthropic | 0.0% | $5.77 | 286 |