Performance by plan.
Every plan runs on the same engine. Higher tiers unlock deeper reasoning, more workers, longer cache, and priority dispatch.
Every plan runs on the same engine. Higher tiers unlock deeper reasoning, more workers, longer cache, and priority dispatch.
How fast documents get embedded and searchable — from 1,000 pages to billions. Scientist plans use OpenAI embeddings with 50× concurrency and deferred semantic enrichment for maximum throughput.
| Ingestion metric | Free | Pro | Business | Enterprise | Scientist | Scientist Team | Scientist Lab |
|---|---|---|---|---|---|---|---|
| 1,000 pages ingest time | ~12 min | ~4 min | ~4 min | ~4 min | ~1 min | ~1 min | ~30 sec |
| Throughput (pages/day) | ~170K pages/day | ~860K pages/day | ~860K pages/day | ~860K pages/day | ~7.2M pages/day | ~7.2M pages/day | ~14M pages/day |
| Max indexed pages | 125K pages | 1M pages | 5M pages | Unlimited | 25M pages | 125M pages | 50M pages |
| Embed concurrency | 3 | 15 | 15 | 15 | 50 | 50 | 100 |
| Embed debounce | 3 min | 2 min | 2 min | 2 min | 30 sec | 30 sec | 15 sec |
| Embedding provider | Cohere | Cohere | Cohere | Cohere | OpenAI | OpenAI | OpenAI |
| Embed batch size | 96 | 96 | 96 | 96 | 2,048 | 2,048 | 4,096 |
| Stream insert + deferred enrichment | Stream only | Stream only | Stream only | Stream only | Both | Both | Both |
Model tier, reasoning depth, agent steps, and verification per plan.
| Capability | Free | Pro | Business | Enterprise | Scientist | Scientist Team | Scientist Lab |
|---|---|---|---|---|---|---|---|
| Price | €0 | €39 | €79 | €149 | €249 | €499 + €199/seat | €4,999 |
| LLM model | Gemini Flash | Gemini Flash / Pro | Gemini Pro / Sonnet | Gemini Pro / Sonnet | Gemini Pro / Sonnet | Gemini Pro / Sonnet | Gemini Pro / Opus |
| Thinking tier | Fast | Balanced | Deep | Deep | Deep | Deep | Deep |
| Max ReAct steps | 3 | 5 | 8 | 8 | 11 | 11 | 14 |
| Max output tokens | 4,096 | 8,192 | 16,000 | 16,000 | 16,000+ | 16,000+ | 32,000+ |
| Step extension (+3) | |||||||
| Reasoning verification | |||||||
| Quality gate | |||||||
| Resonance discovery | |||||||
| Smart templates | |||||||
| Typical write capacity | 1 write | 2-3 writes | 4-5 writes | 4-5 writes | 6-8 writes | 6-8 writes | 6-8 writes |
Workers, cache, priority dispatch, and ingestion capacity per plan.
| Resource | Free | Pro | Business | Enterprise | Scientist | Scientist Team | Scientist Lab |
|---|---|---|---|---|---|---|---|
| Parallel workers | 3 | 5 | 8 | 10 | 32 | 32 | 64 |
| Cache TTL | 24h | 24h | 24h | 24h | 7 days | 7 days | 14 days |
| Cross-tenant cache | |||||||
| Priority dispatch | |||||||
| On-demand boost | +0 | +2 | +5 | +10 | +20 | +20 | +40 |
| Graph retrieval depth | — | 1 level | 2 levels | 2 levels | 3 levels | 3 levels | 4 levels |
| Image OCR | |||||||
| Voice transcription | |||||||
| Embed concurrency | 3 | 15 | 15 | 15 | 50 | 50 | 100 |
| Embed debounce | 3 min | 2 min | 2 min | 2 min | 30 sec | 30 sec | 15 sec |
| Embedding provider | Cohere | Cohere | Cohere | Cohere | OpenAI | OpenAI | OpenAI |
| Embed batch size | 96 | 96 | 96 | 96 | 2,048 | 2,048 | 4,096 |
| Stream insert | |||||||
| Deferred enrichment |
Ephemeral read-only query routes isolate queries from writes, absorb burst traffic, and enable zero-downtime migrations. Higher tiers unlock deeper embedding precision, encryption, and disaster recovery.
| Query route capability | Free | Pro | Business | Enterprise | Scientist | Scientist Team | Scientist Lab |
|---|---|---|---|---|---|---|---|
| Max active query routes | — | 1 | 3 | 5 | 8 | 10 | 12 |
| Query route lifetime (TTL) | — | 2 min | 3 min | 5 min | 10 min | 60 min | 60 min |
| Burst routes (extra) | — | — | +1 | +2 | +3 | +4 | +5 |
| Embedding precision | — | Binary | Summary | Essence | Full | Full | Full |
| Connection pool slots | — | 1 | 2 | 4 | 6 | 10 | 12 |
| Memory per route | 0MB | 8MB | 16MB | 32MB | 64MB | 128MB | 128MB |
| Per-route encryption | |||||||
| Cross-tenant sharing | |||||||
| Snapshot debugging | |||||||
| A/B testing retrieval | |||||||
| Safe re-indexing | |||||||
| Zero-downtime migrations | |||||||
| Disaster recovery failover | |||||||
| Pre-warming routes | |||||||
| Forensic audit replay |
Lift Power, storage, indexing, workspaces, and collaborators per plan.
| Limit | Free | Pro | Business | Enterprise | Scientist | Scientist Team | Scientist Lab |
|---|---|---|---|---|---|---|---|
| Lift Power / billing period | 100K | 2M | 5M | Unlimited | 50M | 50M + 50M/seat | 100M |
| Storage | 250MB | 5GB | 20GB | Unlimited | 100GB | 500GB | 500GB |
| Indexing quota | 250MB | 2GB | 10GB | Unlimited | 50GB | 250GB | 250GB |
| Max workspaces | 1 | 5 | 20 | Unlimited | 10 | 50 | 50 |
| Max collaborators | 0 | 1 | 10 | Unlimited | 5 | 100 | Unlimited |
| SSO / SAML | |||||||
| Audit log | |||||||
| SLA |
How many reasoning actions the AI can take per turn. More steps = deeper analysis. Free gets 3 (one read + one write). Scientist gets 11 (multi-pass analysis with 6-8 writes).
Parallel background jobs per workspace. More workers = faster ingestion. Free: 3. Pro: 5. Business: 8. Scientist: 32 — 10× more parallelism than standard plans.
How long cached answers stay valid. 24h for standard plans. 7 days for Scientist — meaning the same question costs zero compute for a full week after the first ask.
Bypasses the backpressure queue when the system is under load. Scientist plans get +20 on-demand boost — your queries jump ahead of the line.
A second AI model (Gemini 3.5 Pro) independently checks every answer for reasoning errors. Scientist-only. The deepest verification layer available.
If two Scientist users have identical content, the second gets a free cache hit. One computation serves many. Only on Scientist plans.