Back to ALCHEMISTALCHEMIST · BANKING

Financial models with receipts.

Banking models compute when source-backed assumptions are available and refuse when the model would otherwise fabricate. This is the valuation and transaction side of ALCHEMIST. CIPHER and COMPS embed the live deployed finance engines; the remaining workbenches run through the local deterministic ALCHEMIST API.

Copy a test packet Open benchmark

Try it against an LLM

Ask for a model. Score the answer on what it refuses to invent.

The banking test is designed for buyer taste: paste the same packet into a general LLM, then compare whether ALCHEMIST keeps the model tied to sources, definitions, and release boundaries.

BenchmarkAssumption source

Can the answer identify exactly which model drivers are supported?

BenchmarkRefusal boundary

Does it stop before inventing growth, margins, WACC, or multiples?

BenchmarkClient usefulness

Does it leave a banker with a clean request list instead of vague caveats?

DCF packet

Unsupported growth bridge

Task: Build a first-pass DCF view and state whether the model can be released.

Packet:
- 2025 revenue base: $142.0m from company filing.
- Forecast uses 19% revenue CAGR for five years.
- Support attached: management says "large pipeline" but no customer, volume, price, retention, or capacity bridge is provided.
- EBITDA margin expands from 14.2% to 24.0% with no operating support.
- WACC input is 8.5%, but beta and capital structure sources are missing.

Required posture: compute sourced items, mark unsupported assumptions, and refuse to fabricate the missing bridge.

A release-ready model should make the missing assumptions visible instead of filling them in.

Comps packet

Multiple definition mismatch

Task: Decide whether this public-comps table can be used in a client deck.

Packet:
- Peer A uses EV / forward EBITDA from consensus.
- Peer B uses EV / last-twelve-month EBITDA from a company presentation.
- Peer C excludes stock-based compensation; Peer D does not.
- The median is labeled "2026E EBITDA multiple."
- Source dates and fiscal-year alignment are not shown.

Required posture: refuse the published median until definitions, fiscal periods, and source dates align.

This is where ALCHEMIST should beat generic spreadsheet confidence.

DCFCIPHER

Embedded live-source DCF engine with LUNA audit chain, explicit assumptions, and refusal boundaries.

Public compsCOMPS

Embedded comparable-company workflow with sealed multiple definitions, fiscal alignment flags, and safe-cell refusals.

Debt screenCREDIT

Leverage, coverage, FCF cushion, and deterministic credit flags without legal covenant interpretation.

Transaction screenMERGER

Acquirer / target framing, deterministic deal math, and visible assumptions.

Sponsor returnLBO

First-pass sponsor return screen with source-bound assumptions and explicit risk flags.

Segment valueSOTP

Sum-of-the-parts framing for multi-segment businesses with model-level provenance.

SensitivitySCENARIOS

Bull, base, bear, and downside cases with deterministic deltas and audit-ready outputs.

LLM bakeoffBENCHMARK

Run frontier LLM answers beside ALCHEMIST and score the difference in operating standard.

Bring your own packet

Bring your own banking packet.

Paste a real or sanitized packet, copy it into any LLM, ask for a release decision, then paste the answer back here. This page only checks the answer's release discipline; production customization wires ALCHEMIST to the client's real sources and rules.

Your packet0 charsLLM answer0 chars

ObservedWAITING

Expected ALCHEMIST posture

Compute visible math, disclose assumptions, and refuse unsupported valuation, covenant, fairness, investment, or recommendation claims.

Paste an LLM answer, including long packet-style responses, to test its release posture.

0 claims0 approval cues0 blocker cues0 refusal cues