Bill underfunded-alert conformance auditSubscriptions & Billing
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.
65%
GPT-5.5
55%
Claude Opus 4.8
33%
Gemini 3.1 Pro
0%
Gemini 3.5 Flash
Extended metrics
Median Time
Wall-clock minutes, median across trials. N = 10 trials per model.
13m
Gemini 3.1 Pro
6m
Claude Opus 4.8
5m
GPT-5.5
5m
Gemini 3.5 Flash
Tool Calls
Tool invocations, median across trials — counted from the trajectories. N = 10 trials per model.
75
Gemini 3.1 Pro
44
Gemini 3.5 Flash
37
GPT-5.5
37
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_false_funded_blast_radius | 100% | 100% | 80% | 0% |
| answer_spec_underfunded_population | 30% | 10% | 30% | 0% |
| answer_total_spec_shortfall_cents | 30% | 10% | 30% | 0% |
| deliverable_jira_ticket_filed | 100% | 100% | 10% | 0% |
| deliverable_slack_channel_exists | 100% | 100% | 30% | 0% |
| deliverable_slack_message_content | 30% | 10% | 20% | 0% |
| floor_all_connectors | 100% | 100% | 10% | 0% |
| floor_read_source_artifacts | 100% | 100% | 100% | 100% |
| floor_sql_investigation | 100% | 100% | 100% | 100% |
| floor_write_back | 100% | 100% | 10% | 0% |
| trace_all_connectors_and_trap | 30% | 10% | 20% | 0% |
| trace_date_gated_inflow_method | 40% | 30% | 0% | 0% |