1278generated behavioral tests
76.9%official best score
71.4%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) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 71.4% | ok | 913/1278 | $3.61 | 53 | 0.20h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 67.0% | ok | 856/1278 | n/a | 73 | 0.21h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 63.5% | ok | 812/1278 | n/a | 63 | 0.19h | 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
71.4% from 913/1278 ProgramBench-scored tests. Raw public eval statuses: failure: 412, passed: 1172, skipped: 5.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
fb59880c836beval.tests.test_basic_invocation.test_help_specific_command (failure)fb59880c836beval.tests.test_basic_invocation.test_help_for_multiple_commands (failure)fb59880c836beval.tests.test_basic_invocation.test_help_command (failure)fb59880c836beval.tests.test_additional_commands.test_resolve_alias_for_done (failure)fb59880c836beval.tests.test_additional_commands.test_show_next_alias (failure)fb59880c836beval.tests.test_task_modification.test_rm_alias_for_remove (failure)
GPT 5.5 (xhigh) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260518T232006Z
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 | 76.9% | $4.37 | 45 |
| 2 | GPT 5.5 (xhigh) | OpenAI | 71.3% | $11.41 | 107 |
| 3 | Claude Opus 4.7 (xhigh) | Anthropic | 64.0% | $11.13 | 177 |
| 4 | Claude Opus 4.6 | Anthropic | 58.8% | $10.21 | 241 |
| 5 | Claude Sonnet 4.6 | Anthropic | 58.2% | $31.15 | 632 |
| 6 | GPT 5.5 | OpenAI | 55.6% | $1.25 | 16 |
| 7 | Claude Opus 4.7 | Anthropic | 50.7% | $3.06 | 96 |
| 8 | GPT 5.4 | OpenAI | 48.8% | $0.49 | 12 |
| 9 | Claude Haiku 4.5 | Anthropic | 30.8% | $0.69 | 95 |
| 10 | Gemini 3 Flash | 22.6% | $0.31 | 90 | |
| 11 | GPT 5 mini | OpenAI | 7.2% | $0.02 | 18 |
| 12 | Gemini 3.1 Pro | 5.1% | $1.52 | 66 | |
| 13 | GPT 5.4 mini | OpenAI | 1.8% | $0.02 | 7 |