595generated behavioral tests
95.3%official best score
92.3%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 | 92.3% | ok | 549/595 | $4.24 | 75 | 0.23h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 75.6% | ok | 450/595 | n/a | 35 | 0.13h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 75.6% | ok | 450/595 | n/a | 96 | 0.21h | 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
92.3% from 549/595 ProgramBench-scored tests. Raw public eval statuses: failure: 129, passed: 641.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch, terminal rendering mismatch.
d9d4edc20a06eval.tests.test_buffer_operations.TestEdgeCases.test_unicode_emoji_handling (failure)d9d4edc20a06eval.tests.test_buffer_operations.TestTextBufferOperations.test_long_line_wrapping (failure)d9d4edc20a06eval.tests.test_coverage_boost.TestHighlightColors.test_c_file_rendering (failure)d9d4edc20a06eval.tests.test_coverage_boost.TestHighlightColors.test_markdown_file (failure)d9d4edc20a06eval.tests.test_color_output.TestStatusBarColors.test_status_bar_displays (failure)d9d4edc20a06eval.tests.test_buffer_operations.TestTextBufferOperations.test_insert_at_line_end (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 (xhigh) | OpenAI | 95.3% | $9.89 | 82 |
| 2 | Claude Opus 4.6 | Anthropic | 93.3% | $15.07 | 329 |
| 3 | GPT 5.5 (high) | OpenAI | 90.1% | $3.54 | 36 |
| 4 | Claude Sonnet 4.6 | Anthropic | 86.1% | $33.62 | 628 |
| 5 | Claude Opus 4.7 (xhigh) | Anthropic | 82.2% | $2.60 | 89 |
| 6 | GPT 5.5 | OpenAI | 80.5% | $1.09 | 19 |
| 7 | Gemini 3 Flash | 69.7% | $0.35 | 64 | |
| 8 | Gemini 3.1 Pro | 65.9% | $0.84 | 45 | |
| 9 | Claude Opus 4.7 | Anthropic | 40.7% | $0.71 | 46 |
| 10 | Claude Haiku 4.5 | Anthropic | 36.1% | $0.87 | 115 |
| 11 | GPT 5.4 | OpenAI | 34.3% | $0.13 | 9 |
| 12 | GPT 5.4 mini | OpenAI | 26.1% | $0.04 | 8 |
| 13 | GPT 5 mini | OpenAI | 13.6% | $0.01 | 15 |