823generated behavioral tests
99.6%official best score
95.5%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 | 95.5% | ok | 786/823 | $2.48 | 46 | 0.14h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 72.8% | ok | 599/823 | n/a | 66 | 0.19h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 46.2% | ok | 380/823 | n/a | 62 | 0.17h | 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
95.5% from 786/823 ProgramBench-scored tests. Raw public eval statuses: failure: 38, passed: 839.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
34775c99d1b3eval.tests.test_errors.test_duplicate_array_flags_error (failure)34775c99d1b3eval.tests.test_errors.test_duplicate_format_flags_error (failure)adfa09238fcbtests.test_errors.test_cols_overflow_value (failure)adfa09238fcbtests.test_exotic_formats.test_format_flag_empty_string_rejected (failure)adfa09238fcbtests.test_interactions.test_multiple_format_flags_rejected (failure)adfa09238fcbtests.test_interactions.test_multiple_column_flags_rejected (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 | 99.6% | $1.84 | 28 |
| 2 | GPT 5.5 (xhigh) | OpenAI | 98.4% | $7.97 | 87 |
| 3 | Claude Opus 4.7 (xhigh) | Anthropic | 97.3% | $6.72 | 121 |
| 4 | Gemini 3.1 Pro | 91.7% | $1.25 | 84 | |
| 5 | GPT 5.5 | OpenAI | 82.7% | $1.00 | 15 |
| 6 | Claude Sonnet 4.6 | Anthropic | 72.3% | $11.60 | 392 |
| 7 | Claude Opus 4.6 | Anthropic | 71.3% | $5.97 | 227 |
| 8 | GPT 5.4 | OpenAI | 67.2% | $0.19 | 9 |
| 9 | Claude Opus 4.7 | Anthropic | 66.3% | $2.75 | 89 |
| 10 | Gemini 3 Flash | 66.1% | $0.34 | 124 | |
| 11 | GPT 5.4 mini | OpenAI | 36.1% | $0.03 | 13 |
| 12 | Claude Haiku 4.5 | Anthropic | 28.7% | $0.63 | 114 |
| 13 | GPT 5 mini | OpenAI | 4.4% | $0.03 | 21 |