Skip to main content

Pricing

Benchmarked against: Anthropic โ€” Pricing Rule source: Company Constitution ยง6, Cost Awareness governance rule Captain insight: Paying 20x Max Plan + large API deposits โ€” every token has real cost

SuperPortia's cost structure spans Claude subscription plans, low-cost API engines, and infrastructure services. Understanding these costs is essential for engine selection and operational efficiency.


Claude pricingโ€‹

Subscription plansโ€‹

ResourceBillingShared withNotes
All Models quota (Opus, Sonnet, Haiku)Max Plan monthlyclaude.ai Chat + Claude Code CLIChat (ๅฐ่ฅฟ) and Code (ๅฐๅ…‹) share this quota
Sonnet-only quotaSeparate monthly allowanceclaude.ai Chat + Claude Code CLIIndependent โ€” does not consume All Models
API credits (extra usage / LiteLLM)Per-tokenDirect API calls onlyKicks in when quota exceeded

Critical insight: All Models quota is shared between claude.ai (Chat tab = ๅฐ่ฅฟ) and Claude Code (Code tab = ๅฐๅ…‹). Heavy CLI usage directly reduces Chat availability, and vice versa.

Claude per-token rates (API / extra usage)โ€‹

ModelInputOutput
Opus 4.6Most expensiveMost expensive
Sonnet 4.6ModerateModerate
Haiku 4.5Cheapest ClaudeCheapest Claude

Exact per-token pricing follows Anthropic's published rates. Check anthropic.com/pricing for current numbers.


Low-cost engine pricingโ€‹

EngineDefault ModelCost per requestMonthly estimate (100 req/day)
GroqLlama 3.3 70BFree$0
Groq SearchCompoundFree$0
Gemini2.5 Flash~$0.003~$9
Gemini Search+ Google Grounding~$0.014~$42
DeepSeekR1 / V3Cents~$5-10
MistralLatestCents~$5-10
ZhipuGLM-5Cents~$5-10
IngestMTAAA PipelineFree (internal)$0

Infrastructure pricingโ€‹

All infrastructure currently operates within free tiers:

ServiceFree tierCurrent usageOverage cost
Cloudflare Workers100K requests/dayWell within$0.50/million
Cloudflare D15M reads/day, 100K writes/dayWell within$0.001/million reads
Cloudflare Vectorize30M queries/monthWell withinUsage-based
Cloudflare R210GB storage, 1M reads/monthWell within$0.015/GB/month
SupabaseFree tier / Pro planVaries by projectPlan-based

Cost comparison: engine selection impactโ€‹

The same task at different engine levels:

TaskFree (Groq)Cheap (Gemini)Standard (Sonnet)Premium (Opus)
Web search$0$0.014$0.10+$0.50+
Text summary$0$0.003$0.05+$0.30+
Code generationN/AN/A$0.10-0.50$0.50-2.00
File operationsN/AN/A$0.10-0.50$0.50-2.00
WO dispatch$0$0.01$0.50-1.00$1.00-2.00

Key takeaway: Delegating research to Groq/Gemini instead of using Opus directly saves 10-100x per query.


Cost optimization strategiesโ€‹

StrategySavingsHow
UB-first searchMassivesearch_brain() before any external search โ€” answer may already be ingested
Delegate research10-100xUse intel_search (Groq free) or search_web instead of Opus WebSearch
Batch patrolLinearRun patrol once for a domain, not individual searches
Engine matchingVariableMatch engine to task importance per selection guide
Progressive DisclosureToken savingsLoad skills on demand, not at startup
Concise promptsToken savingsSlim rules, skills load only when invoked

Monitoring and trackingโ€‹

# Check engine availability and API key status
list_models()

# Check UB volume (indicates ingestion costs)
get_stats()

# Check WO history (each WO has engine + actual_hours)
list_work_orders(include_completed=True)

Monitoring gaps (planned)โ€‹

FeatureStatus
Per-session token counterPlanned
Per-WO cost estimationPlanned
Monthly cost dashboardPlanned
Budget alertsPlanned
Cost-per-capability trackingPlanned

PageRelationship
Choosing an EngineSelection framework
Usage & CostAdmin tracking tools
Cost AwarenessGovernance rule
Engine OverviewFull engine catalog