Token usage
916,720,195 gross tokens (912,134,236 input + 4,585,959 output) across 821 recorded messages from 262 agents (316 in project scope), snapshot 2026-11-06 08:20 PST (2026-11-06T16:20:14Z UTC).
| Scope | Input | Output | Gross |
|---|---|---|---|
| design chat tree (main chat + its subagents) | 676,515,174 | 3,325,646 | 679,840,820 |
| checkpoint cron (pdlvds-progress) | 235,619,062 | 1,260,313 | 236,879,375 |
Records are the runtime’s per-message token-usage rows: each row aggregates every model request made while producing one message — the full context is re-sent on every request, so input dominates and the totals are a work meter, not a content size. Main-chat rows are true per-message values; a subagent appears as a single whole-run record timestamped at its completion. Agents still running have no rows yet, so the snapshot understates the final total. Full method, the exact SQL and all caveats: metrics/README.md.
Board progress against cumulative tokens: progress_vs_tokens.svg (unconnected nets and clearance violations from the audit file as committed at each revision).
Cost summary
What this workload would cost at each provider’s published API list prices, computed from the same per-message token records (method and price basis: metrics/README.md).
The rates are constants — US dollars per million tokens, standard pay-as-you-go list prices:
| Provider | Model | Input | Cached input | Output |
|---|---|---|---|---|
| Meta | Muse Wonder 1.0 | $4.00 | $0.20 | $20.00 |
| Meta | Muse Spark 1.3 | $1.25 | $0.15 | $4.25 |
| Anthropic | Opus 5.5 | $4.00 | $0.20 | $20.00 |
| Anthropic | Sonnet 5.5 | $2.00 | $0.10 | $10.00 |
| OpenAI | GPT-6 Astra | $10.00 | $1.00 | $50.00 |
| OpenAI | GPT-6.1 Sol | $2.00 | $0.10 | $10.00 |
Cost at list prices — each model’s share priced at its own tier (the main chat and the checkpoint cron ran on the flagship-tier model, the subagents on the step-down-tier model):
| Scope | Meta | Anthropic | OpenAI |
|---|---|---|---|
| Main chat | $1,125.42 | $1,125.42 | $2,813.55 |
| Subagents | $509.93 | $823.58 | $823.58 |
| Checkpoint cron | $967.69 | $967.69 | $2,419.21 |
| Total | $2,603.04 | $2,916.69 | $6,056.34 |
Two whole-volume variants. Upper bound (all flagship): every token priced at the flagship tier, ignoring the tier split. Cache-adjusted (measured share): repeated input billed at the cached-input rate, using the measured repeated share of input instead of the standard rate.
| Meta | Anthropic | OpenAI | |
|---|---|---|---|
| All-cached (tier-matched) | $216.38 | $210.18 | $685.67 |
| Cache-adjusted (pessimistic estimate) | $1,722.68 | $1,880.37 | $4,108.76 |
| Cache-adjusted (optimistic estimate) | $300.70 | $300.07 | $880.48 |