855generated behavioral tests
63.4%official best score
59.9%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 | 59.9% | ok | 512/855 | n/a | 47 | 0.14h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 54.0% | ok | 462/855 | $6.58 | 88 | 0.28h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 52.2% | ok | 446/855 | n/a | 71 | 0.18h | 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) · 20260518T232006Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260517T094700Z
54.0% from 462/855 ProgramBench-scored tests. Raw public eval statuses: failure: 475, passed: 672, skipped: 20.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, terminal rendering mismatch.
95441a0fda1eeval.tests.test_basic_cli.test_conflicting_path_options (failure)95441a0fda1eeval.tests.test_daemon_comprehensive.test_daemon_exit_immediately_clean_exit (failure)95441a0fda1eeval.tests.test_daemon_comprehensive.test_daemon_exit_immediately_missing_instruction (failure)95441a0fda1eeval.tests.test_daemon_comprehensive.test_daemon_malformed_json_empty_string (failure)95441a0fda1eeval.tests.test_tui_advanced.test_tui_stash_operations (failure)95441a0fda1eeval.tests.test_daemon_comprehensive.test_daemon_malformed_json_invalid_syntax (failure)
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 (high) | OpenAI | 63.4% | $3.21 | 51 |
| 2 | GPT 5.5 (xhigh) | OpenAI | 62.2% | $7.85 | 87 |
| 3 | Claude Opus 4.6 | Anthropic | 56.4% | $9.52 | 188 |
| 4 | Claude Sonnet 4.6 | Anthropic | 51.2% | $8.11 | 206 |
| 5 | Gemini 3 Flash | 46.1% | $0.16 | 41 | |
| 6 | Claude Opus 4.7 (xhigh) | Anthropic | 39.2% | $8.39 | 168 |
| 7 | Claude Opus 4.7 | Anthropic | 36.7% | $0.93 | 49 |
| 8 | Gemini 3.1 Pro | 32.7% | $1.29 | 87 | |
| 9 | GPT 5.4 | OpenAI | 28.3% | $0.39 | 11 |
| 10 | Claude Haiku 4.5 | Anthropic | 26.8% | $0.31 | 65 |
| 11 | GPT 5 mini | OpenAI | 24.3% | $0.02 | 15 |
| 12 | GPT 5.4 mini | OpenAI | 23.0% | $0.04 | 118 |
| 13 | GPT 5.5 | OpenAI | 3.3% | $0.92 | 15 |