Fast is a feature.
Every benchmark and service level target that runs Falcon — publicly stated so you can hold us to them.
Instant, everywhere.
Core Web Vitals engineered for the 75th percentile — not lab conditions.
Initial page load
Under 2 seconds — cold, from the edge, worldwide.
Largest Contentful Paint
LCP under 2.5s at the 75th percentile across markets.
First Contentful Paint
FCP under 1.5s so the app feels alive immediately.
Lighthouse Score
95+ across performance, accessibility, best-practices and SEO.
Time to Interactive
Under 3 seconds — click as soon as you see it.
Streams before you blink.
Falcon Router picks the fastest capable model per request.
Stream start
First token in 1–2 seconds for frontier models via Falcon Router.
Response completion
Average end-to-end AI response under 15 seconds.
Token usage display
Live cost + token counter updates under 500ms.
Conversation load
Full thread hydrated in under 1 second.
New chat creation
Under 300ms from click to first prompt-ready state.
Grounded in milliseconds.
Hybrid retrieval and reranking under a second — even at corpus scale.
Upload acknowledgement
Document accepted and queued under 2 seconds.
Embedding pipeline
Chunking + embedding starts within 5 seconds of upload.
Vector search
Semantic retrieval under 300ms across large corpora.
Reranking
Cross-encoder rerank under 700ms for top-k precision.
End-to-end RAG
Retrieval + grounding under 5s before streaming begins.
Deck to done.
Every format your team ships — generated in seconds.
Markdown
Generated and returned in under 5 seconds.
DOCX
Formatted Word documents in under 10 seconds.
XLSX
Multi-sheet workbooks with formulas in under 12 seconds.
PPTX
Full decks with layout and imagery in under 15 seconds.
PDF export
Print-ready PDF export in under 10 seconds.
Realtime, not near-time.
Sub-second sync for messages, presence and shared documents.
Realtime message sync
Cross-client delivery under 500ms globally.
Presence updates
Join, leave and cursor presence under 1 second.
Shared document updates
Collaborative edits propagate under 1 second.
Predictable latency, always.
Fast gateway, real-time rate limits, delivery-guaranteed webhooks.
Gateway latency
Under 200ms excluding upstream model time.
Availability
99.9% platform availability target for API traffic.
Rate limiting
Real-time, per-key and per-workspace enforcement.
Webhook processing
Delivery attempted within 5 seconds of the source event.
Built to fail over.
Global availability targets with disciplined RTO and RPO.
Global uptime
99.9% SLA across regions and edges.
Disaster recovery — RTO
Recovery time objective under 4 hours for critical services.
Disaster recovery — RPO
Recovery point objective under 15 minutes of data loss.
Automatic failover
Health-checked failover for gateway, auth, storage and models.
Architected for millions.
Horizontal scale on Cloudflare's edge and Supabase primitives.
Registered users
Architected for 100,000+ registered accounts per tenant.
Concurrent users
10,000+ concurrent sessions per workspace region.
Monthly AI requests
Millions of requests per month, horizontally scaled.
Elastic infrastructure
Cloudflare edge + Supabase primitives scale on demand.
The numbers we watch.
Product, reliability and business KPIs — reviewed continuously.
Error rate
Under 1% across all services — tracked per route and per region.
Failed AI requests
Under 0.5% including model outages, retries and fallbacks.
Payment success rate
Over 99% for card, wallet and enterprise invoicing.
Customer satisfaction
CSAT over 90% across product and support surfaces.
Monthly churn
Under 8% — actively managed with retention automation.
Commitments in writing.
Uptime credits, priority support and proactive incident communication.
Uptime SLA
99.9% contractual uptime with service credits for enterprise plans.
Priority Support
Dedicated response times for Sev-1 through Sev-4 incidents.
Incident notifications
Proactive alerts through email, Slack and webhook channels.
Dedicated monitoring
Named contacts, escalation paths and on-call rotations.