EU AI ACTto the 2 December 2027 Annex III enforcement deadline.Check your tier →
Home · Models · Lekha 2.5
Sovereign model family · Banking, financial services & insurance

Onboarding, KYC, fraud signals and claims — inside the bank's perimeter.

Lekha is the BFSI family. It reads the documents a bank or insurer runs on — identity, income, policy, claim — and drives straight-through processing without a token leaving the perimeter. Deployed on bank-controlled infrastructure with the audit trail regulators expect.

Live & Stable
Status
v2.5
Current version, frozen per deployment
4-bit
Quantised for your hardware
0
Outbound API calls
Live & Stable
Lk
MindMap Lekha 2.5
Lekha — the Sanskrit word for a written record or ledger.
Tuned for

Onboarding · KYC · fraud signals · claims straight-through processing

Onboarding & KYC

Identity, address and income documents verified and extracted end-to-end; exceptions routed, everything else straight through.

Claims straight-through processing

First-notice-of-loss to settlement recommendation on the carrier's own policy wording and historical decisions.

Fraud signals

Document-level tampering and inconsistency signals surfaced to the fraud team before the case is booked.

Regulatory Q&A

Branch and compliance staff assistants grounded on the institution's own circulars, manuals and product terms.

Runs where the data lives

Bank perimeter or private cloud — air-gap capable, with every prompt and retrieval streamed to the bank's SIEM.

In production

Live and stable at a Gulf digital bank (customer onboarding and KYC at 98% straight-through) and a US insurance carrier.

References are anonymised by default; named references on request under NDA.

Governance

What comes with every Lekha deployment

Provenance register

Base model, version and licence recorded for every deployment — the first artefact an auditor or a procurement team asks for.

Golden set per domain

Every family carries a curated evaluation set drawn from the document types its clients run on. Releases are gated on it.

Blind human + AI-judge scoring

Outputs scored by domain reviewers without knowing which model produced them, cross-checked by an AI judge. Raw runs available on request.

Frozen in production

The model that served your customers on Friday is the model serving them on Monday. No silent upgrades, no vendor deprecations, no pricing changes.

Plainly: Lekha is a domain fine-tune of a permissively-licensed open-weight base, not a from-scratch foundation model. The base, its version and its licence are recorded in the provenance register for every deployment.
Evaluate before you commit

Run Lekha on your own documents.

Send 20–50 representative documents or transcripts — anonymised is fine — and the acceptance rules your team uses today. We run Lekha 2.5 on infrastructure you control, score it against your criteria, and hand back the results and the harness. Two weeks, fixed scope, no production data.

An engineer replies within one business day. Evaluations run on infrastructure you control; we never ask for production data.

Talk to the product team