ACH-debit holds released before scheduled settlementPayments & Transfers
How each model performed on this task. Every chart states its N, and the reward carries a 95% confidence interval once there is more than one trial to measure.
Core metrics
Check Pass %
Mean share of the task's checks passing, averaged over trials. N = 10 trials per model.
77%
GPT-5.5
52%
Gemini 3.1 Pro
47%
Claude Opus 4.8
3%
Gemini 3.5 Flash
Extended metrics
Median Time
Wall-clock minutes, median across trials. N = 10 trials per model.
7m
Gemini 3.1 Pro
5m
Claude Opus 4.8
4m
Gemini 3.5 Flash
4m
GPT-5.5
Tool Calls
Tool invocations, median across trials — counted from the trajectories. N = 10 trials per model.
62
GPT-5.5
54
Gemini 3.1 Pro
33
Gemini 3.5 Flash
32
Claude Opus 4.8
Verifier × model pass rate
How often each verifier passed, per model. Green is passing, red is failing — a red band across a row is a verifier nothing clears.
| Verifier | GPT-5.5n=10 | Claude Opus 4.8n=10 | Gemini 3.1 Pron=10 | Gemini 3.5 Flashn=10 |
|---|---|---|---|---|
| checks (12) | ||||
| answer_early_release_count | 70% | 20% | 40% | 0% |
| answer_exposure_usd | 70% | 20% | 40% | 0% |
| answer_timing_and_service_split | 70% | 20% | 40% | 0% |
| deliverable_jira_ticket_filed | 90% | 100% | 60% | 10% |
| deliverable_slack_channel_exists | 90% | 100% | 90% | 10% |
| deliverable_slack_message_content | 70% | 20% | 40% | 0% |
| floor_all_connectors | 90% | 100% | 90% | 10% |
| floor_read_source_confluence | 100% | 100% | 100% | 100% |
| floor_sql_investigation | 100% | 100% | 100% | 100% |
| floor_write_back | 90% | 100% | 90% | 10% |
| trace_all_connectors_and_trap | 70% | 20% | 90% | 0% |
| trace_hold_release_method | 50% | 0% | 20% | 0% |