Models and cost
What each agent runs on in production, what has been measured, and what a run should cost.
What the agents run on
| agent | model | $ / M input | $ / M output |
|---|---|---|---|
| steward | Claude Fable 5.1 claude-fable-5-1 | 10.00 | 50.00 |
| solver | Claude Fable 5.1 claude-fable-5-1 | 10.00 | 50.00 |
| steward strong | Claude Fable 5.1 claude-fable-5-1 | 10.00 | 50.00 |
| curator | Claude Fable 5.1 claude-fable-5-1 | 10.00 | 50.00 |
| audit | Claude Fable 5.1 claude-fable-5-1 | 10.00 | 50.00 |
| arbitration | Claude Fable 5.1 claude-fable-5-1 | 10.00 | 50.00 |
| extractor | Claude Fable 5.1 claude-fable-5-1 | 10.00 | 50.00 |
| matcher | GLM 5.3 Flash z-ai/glm-5.3-flash | provider-priced | provider-priced |
| tagger | GLM 5.3 Flash z-ai/glm-5.3-flash | provider-priced | provider-priced |
| judge (evals only) | Claude Sonnet 5 claude-sonnet-5 | 3.00 | 15.00 |
List rates from pricing.ts, the table the system bills through. Pins from api-stack.ts. --profile=production applies exactly these.
Detail
What has been measured
| what | setup | cost |
|---|---|---|
| golden pairs, 30 decisions | DeepSeek V4 Flash | $0.06 |
| blackholes ingest, 11 Steward runs, capped | Sonnet Steward, 12 iterations, Curator off | $11.29 |
| judge, 13 verdicts | Claude Sonnet 5 | $0.60 |
What a run should costestimate
| cluster | documents | Steward runs | per run |
|---|---|---|---|
blackholes | 4 | ≈28 | $80 – 200 |
eggs | 3 | ≈20 | $60 – 140 |
lableak | 5 | ≈35 | $100 – 250 |
lethalities | 11 | 100 – 150 | $400 – 900 |
Basis, as of 2026-09-04: the Steward is nearly the whole cost, one per extracted claim (about seven per document), each a tool-using loop over a prompt that carries the entire constitution. A Sonnet Steward capped at twelve iterations cost about $1 a run; production uses Claude Fable 5.1 at 3.3× the rate with a 200-iteration budget, so $2 – 6 per Steward run, plus Curator sweeps. The first metered production run replaces this table.
Spending limits
Set in .env; 0 means unlimited. The production profile sets models only, not these.
STEWARD_MAX_RUNS: total Steward runs in a run. The main cap.STEWARD_MAX_ITERATIONS: tool calls within one Steward run. 200 by default; lower it only for smoke tests.CURATOR_MAX_RUNS,CURATOR_SWEEP_RATE: Curator sweeps; rate 0 disables.LLM_DAILY_TOKEN_LIMIT,LLM_HOURLY_TOKEN_LIMIT: the circuit breaker; the run stops cleanly when hit.
Every run prints its exact metered cost at the end.