1978generated behavioral tests
72.5%official best score
67.4%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) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 67.4% | ok | 1334/1978 | n/a | 50 | 0.17h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 66.0% | ok | 1305/1978 | n/a | 45 | 0.16h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 19.5% | ok | 385/1978 | $4.23 | 69 | 0.24h | 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) · 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.
GPT 5.5 (xhigh) · 20260517T094700Z
19.5% from 385/1978 ProgramBench-scored tests. Raw public eval statuses: failure: 1640, passed: 420, skipped: 2.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
008e5eb37c59eval.tests.test_tcc.test_tests2_42_to_50 (failure)008e5eb37c59eval.tests.test_tcc.test_ex1_hello_world (failure)008e5eb37c59eval.tests.test_tcc.test_tests2_21 (failure)008e5eb37c59eval.tests.test_tcc.test_tests2_do_while (failure)008e5eb37c59eval.tests.test_tcc.test_run_mode (failure)008e5eb37c59eval.tests.test_tcc.test_complex_preprocessor (failure)
Official ProgramBench Results by Model
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | GPT 5.5 (xhigh) | OpenAI | 72.5% | $6.30 | 65 |
| 2 | GPT 5.5 (high) | OpenAI | 17.8% | $3.43 | 36 |
| 3 | GPT 5.5 | OpenAI | 16.0% | $1.29 | 17 |
| 4 | GPT 5.4 | OpenAI | 12.8% | $0.43 | 12 |
| 5 | Gemini 3.1 Pro | 9.7% | $1.52 | 108 | |
| 6 | Claude Sonnet 4.6 | Anthropic | 9.3% | $100.12 | 879 |
| 7 | Claude Opus 4.6 | Anthropic | 8.0% | $12.01 | 189 |
| 8 | Gemini 3 Flash | 4.8% | $0.16 | 50 | |
| 9 | Claude Opus 4.7 (xhigh) | Anthropic | 3.7% | $0.75 | 22 |
| 10 | Claude Opus 4.7 | Anthropic | 3.3% | $0.45 | 22 |
| 11 | GPT 5.4 mini | OpenAI | 2.2% | $0.02 | 7 |
| 12 | GPT 5 mini | OpenAI | 1.6% | $0.02 | 12 |
| 13 | Claude Haiku 4.5 | Anthropic | 0.0% | $0.40 | 108 |