1004generated behavioral tests
85.7%official best score
75.0%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 | 75.0% | ok | 753/1004 | n/a | 81 | 0.26h | not exported |
| 2 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 53.0% | ok | 532/1004 | $5.18 | 85 | 0.22h | manifest · eval json · eval summary · usage audit |
| 3 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 43.8% | ok | 440/1004 | n/a | 40 | 0.14h | 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
53.0% from 532/1004 ProgramBench-scored tests. Raw public eval statuses: failure: 44, not_run: 445, passed: 703.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
7c5468344bbatests.test_combinations.test_break_flag_adds_newlines_between_files (not_run)7c5468344bbatests.test_combinations.test_column_with_vimgrep_redundancy (not_run)7c5468344bbatests.test_combinations.test_context_before_and_after_with_multiple_matches (not_run)7c5468344bbatests.test_combinations.test_context_with_only_matching_shows_context_lines (not_run)7c5468344bbatests.test_combinations.test_count_shows_match_count_per_file (not_run)7c5468344bbatests.test_combinations.test_depth_flag_limits_directory_traversal (not_run)
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 | 85.7% | $9.59 | 98 |
| 2 | Claude Opus 4.7 (xhigh) | Anthropic | 78.3% | $6.00 | 113 |
| 3 | GPT 5.5 (high) | OpenAI | 76.6% | $4.66 | 52 |
| 4 | Claude Opus 4.6 | Anthropic | 70.1% | $7.38 | 195 |
| 5 | GPT 5.5 | OpenAI | 59.6% | $1.54 | 20 |
| 6 | Claude Sonnet 4.6 | Anthropic | 59.3% | $13.76 | 443 |
| 7 | Gemini 3.1 Pro | 55.7% | $1.38 | 103 | |
| 8 | GPT 5.4 | OpenAI | 51.7% | $0.20 | 6 |
| 9 | Claude Haiku 4.5 | Anthropic | 38.9% | $0.88 | 144 |
| 10 | Gemini 3 Flash | 26.6% | $0.33 | 91 | |
| 11 | GPT 5 mini | OpenAI | 10.6% | $0.01 | 9 |
| 12 | GPT 5.4 mini | OpenAI | 9.6% | $0.39 | 323 |
| 13 | Claude Opus 4.7 | Anthropic | 1.5% | $3.72 | 107 |