Model economics / detail
claude-opusC
Dispatch tier: Expensive (complexity class this model is routed for)
One run, lost by the infra, then left 'running' for 17 days — a 24,482-minute monument to nobody checking on it. Zero evidence of skill or failure; C by presumption of innocence, not merit.
The measured record
What the ledger shows
Fleet-ledger aggregates for this model only — run counts, outcomes, spend, and token appetite. Missing telemetry reads “Not observed”, never zero.
Agent runs
1
Success rate
0%
Metered spend
Not observed
cost-blind: no run was metered
Cost / metered run
Not observed
Tokens in
Not observed
Tokens out
Not observed
Tokens in / run
Not observed
Tokens out / run
Not observed
Outcomes
Status breakdown
Every run ends in exactly one status. The bar is the whole record, to scale.
- lost: 1 of 1 runs (100.0% of total)
Run duration
How long the runs take
Wall-clock distribution across measured runs, from the fastest exit to the longest grind.
Min
408.0h
p25
408.0h
Median
408.0h
p75
408.0h
p90
408.0h
p99
408.0h
Max
408.0h
Mean
408.0h
Duration measured on 1 of 1 runs.
Commentary
Idiosyncrasies
What stands out in this model's numbers — shape, appetite, and failure habits.
- One run, lost by the infra, with a recorded duration of 1,468,894 seconds — 17 days. That is a bookkeeping zombie, not a runtime.
- The 17-day figure will poison any duration average it is allowed near.
Commentary
Lessons learned
Practical routing and operations takeaways, grounded in the same record.
- Nothing to learn yet beyond: fix the router — and add a watchdog so a lost run cannot impersonate a 17-day job.