AI enginesFast, Core, Expert, Ultra

One assistant. Four depths of work.

Every LumenQube editor runs the same assistant. What you choose is how far it goes to be right — one fast pass, a build it checks, a build it proves a second way, or an engine that interrogates your data before it trusts it.

Download freeSee the same job, four ways

Every plan reaches Lumen Core through Automatic routing. Pinning an engine starts on Plus.

Engines
Four depthsFast, Core, Expert, Ultra
Choosing
AutomaticOr pin one, from Plus up
Allowance
One per monthDeeper runs use more of it
Extended thinking
From Pro upDeeper reasoning on Expert and Ultra

01What it does

Depth is a shape, not a slider.

The engine changes how many of the same five steps a run gets, how many times it checks itself, and how much room it has in a single turn.

01The shape of a run

Every run is the same five steps.

Gather, plan, create, review, enhance. Fast takes the short road. Deeper engines add the plan, then loop review and enhance until the result stops changing — reading the finished work back out of the document, not from memory.

  • Nothing touches your file until the plan exists
  • Review reads the result back from the document
  • Enhance fixes it and re-derives the headline numbers
01Gather02Plan03Create04Review05EnhanceSteps
FastAnswers directly skippedskippedskipped 12
CoreChecks once oncefixes 20
ExpertChecks up to twice up to ×2re-derives 30
UltraChecks the data first + sourceup to ×3re-derives 42
Step budgets shown are LumenSheets’. Every editor sets its own, and a deeper engine always gets more.

02The same four depths, in every editor

One assistant across PDFs, sheets, documents, slides, design, notes and forms. Each editor decides what each depth is for — and what it is allowed to check.

FastCoreExpertUltra
LumenPDFA 90-page agreement Explain a clause, pull the datesA structured briefCompare two versions clause by clauseHold the whole stack in mind
LumenSheetsA quarter close Explain a formula, clean a columnA working model with a chartTrace a broken model to its cellLarge models checked at source
LumenDocsA first draft Tighten and fix the toneA full draft from notesAudit for gaps and unsupported claimsAmbiguous briefs, resolved
LumenDesignA composed layout Captions, a headline, layersA real layout in your brandJudges and fixes its own outputArt direction, a coherent system

03What “checks its own work” means

It reads the result back from the real document — the sheet’s values, the rendered page, the board as drawn — and fixes what does not match.

01Writes the resultSeptember · $451,206 · −7.3%—
02Reads it back out of the sheetSums the invoices behind cell B6—
03Finds the gapINV-3308 and INV-3311 appear twice · $12,294Mismatch
04Fixes and re-derivesSeptember · $438,912 · −9.9%Clean

04One allowance

Every account gets a monthly allowance. A deeper engine takes more steps, so it uses more of it — use Fast for the many small asks and save depth for the work that earns it.

This month’s allowance71% left
Fast62 small asksCore14 documentsExpert4 models checkedUltra1 hard brief
Cap any single run — at the ceiling it stops and reports.

05Which plan reaches which engine

Free starts on Lumen Fast, and Automatic steps real work up to Lumen Core. From Plus you can pin an engine; Pro adds Expert on demand, Max runs on Expert and adds Ultra, and the Ultra plan runs on Ultra.

Extended thinking from Pro up

EngineFreePlusProMaxUltra
Default engineFastCoreCoreExpertUltra
Lumen Fast
Lumen CoreAutomatic
Lumen Expert——On demand
Lumen Ultra———On demandAlways
Extended thinking——
Pin an engine—

02LumenAgent + engines

Depth matters most when nobody is watching.

LumenAgent runs on your plan’s engine. In a long autonomous run every decision compounds, so the deeper engines check the work before they hand it back — and say what they changed.

Gathers
The files and the ledger behind the ask
Reviews
The finished result, read back from the document
Fixes
What review found, then re-derives the totals
Reports
What it changed, and why

03The full toolkit

Pick the depth, not the model.

Four engines, one assistant. The difference is how far each one goes before it calls the work finished.

Lumen Fast

The dozens of small asks.

  • Answers almost immediately
  • Uses the least of your month
  • No self-check
  • 12 tool steps in LumenSheets

Lumen Core

Real document work.

  • Holds a long file in mind
  • Follows multi-part instructions
  • Checks once before finishing
  • 20 steps · every plan

Lumen Expert

When wrong is expensive.

  • Thinks before answering
  • Proves the answer a second way
  • Checks up to twice
  • 30 steps · Pro on demand

Lumen Ultra

The hardest work you have.

  • Our most capable engine
  • Checks the data before writing
  • Extended thinking
  • 42 steps · Max on demand, Ultra always

04Questions

Common questions

Do I have to pick an engine every time?

No. Each plan has a default, and Automatic steps real work up to Lumen Core on every plan. From Plus, the picker in each editor’s AI panel is there for the task that deserves more depth.

What does “checks its own work” actually mean?

It reads the result back from the real document — sheet values, the rendered page, the board as drawn — and fixes what does not match. Fast skips this; Core does it once; Expert and Ultra repeat it and re-derive the headline numbers.

Which engines does my plan include?

Free starts on Lumen Fast and reaches Lumen Core through Automatic routing. Plus defaults to Core and lets you pick. Pro defaults to Core with Lumen Expert on demand. Max defaults to Expert with Lumen Ultra on demand, and the Ultra plan runs on Ultra. Extended thinking comes with Pro, Max and Ultra.

Does a deeper engine cost more?

It uses more of the same monthly allowance, because it takes more steps and checks more. On paid plans one day can use up to 30% of the month’s allowance, and you can cap what a single run may spend.

Which engine does LumenAgent use?

Your plan’s engine. Depth matters most there, because in a long run every decision compounds.

Are my documents used to train these models?

LumenQube does not use your content to train a generalized AI model as a product purpose. Content is sent only when you request an AI feature, to the selected processor or eligible fallback needed for that request. A provider’s own data use, retention and permitted security or legal processing depend on the service and our account terms and settings. See Privacy Policy §7.

Every engine, in one app.

Download free and start with Lumen Fast and Lumen Core and a monthly allowance. Move up when the work asks for it.

Download freeCompare plans

Every editor is free, with no trial timer.