2926generated behavioral tests
74.7%official best score
69.5%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 | 69.5% | ok | 2033/2926 | $9.60 | 131 | 0.45h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 51.2% | ok | 1499/2926 | n/a | 75 | 0.26h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 2.1% | ok | 60/2926 | n/a | 58 | 0.22h | 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
69.5% from 2033/2926 ProgramBench-scored tests. Raw public eval statuses: failure: 1307, passed: 2209, skipped: 21.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch.
c9670de85f04tests.test_comprehensive_scenarios.test_string_manipulation_chain (failure)c9670de85f04tests.test_base64_functions.test_base64_encode_urlsafe (failure)c9670de85f04tests.test_advanced_features.test_nested_template (failure)c9670de85f04tests.test_additional_functions.test_math_seq_with_step (failure)c9670de85f04tests.test_additional_functions.test_multiple_delimiters_in_one_template (failure)c9670de85f04tests.test_comprehensive_scenarios.test_function_composition (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 | Claude Sonnet 4.6 | Anthropic | 74.7% | $22.13 | 450 |
| 2 | GPT 5.5 (high) | OpenAI | 71.8% | $5.18 | 46 |
| 3 | Claude Opus 4.6 | Anthropic | 70.9% | $14.37 | 258 |
| 4 | Claude Opus 4.7 (xhigh) | Anthropic | 57.6% | $6.97 | 133 |
| 5 | GPT 5.5 (xhigh) | OpenAI | 54.7% | $8.30 | 70 |
| 6 | GPT 5.5 | OpenAI | 52.1% | $1.56 | 22 |
| 7 | Claude Opus 4.7 | Anthropic | 41.1% | $4.69 | 89 |
| 8 | GPT 5.4 | OpenAI | 19.3% | $0.46 | 12 |
| 9 | Gemini 3 Flash | 17.1% | $0.24 | 89 | |
| 10 | Claude Haiku 4.5 | Anthropic | 14.4% | $0.48 | 98 |
| 11 | GPT 5 mini | OpenAI | 10.4% | $0.02 | 17 |
| 12 | GPT 5.4 mini | OpenAI | 8.6% | $0.05 | 10 |
| 13 | Gemini 3.1 Pro | 0.0% | $0.90 | 86 |