Model Development

Models you can trust

We develop robust risk models and carry out independent validation of models built by others — including artificial intelligence models.

Validation report

Illustrative example

Conceptual soundness

Adequate

Data quality

Adequate

Replication and implementation

Adequate

Performance and backtesting

With reservations

Documentation and governance

Needs improvement

✓

Fit for use, with conditions

3 recommendations · 90-day remediation window

Português

What we do

From development to monitoring

01

Model development

Credit, provisioning, liquidity, pricing and capital models, built with full documentation from day one.

02

Independent validation

Critical, independent review of internal, vendor or consultant-built models, with clear conclusions.

03

AI and machine learning models

Validation of explainability, bias, stability and robustness of AI models.

04

Model risk management

Model inventory, materiality tiering, policies and model lifecycle.

05

Performance monitoring

Performance indicators, periodic backtesting and deterioration alerts.

06

Remediation

Addressing audit, validation or supervisory findings, with a plan and evidence.

Independent validation

A second technical opinion, free of conflicts of interest

Whoever builds a model should not be the one who validates it. Independent validation gives the board, auditors and supervisors confidence that the model is fit for its intended use — and transparency about its limitations.

What we review

Full validation scope

We assess the model end to end, proportionate to its materiality.

✓

Conceptual soundness and assumptions

✓

Data quality and representativeness

✓

Independent replication of results

✓

Performance, backtesting and benchmarking

What you receive

Actionable conclusions

A clear report, with findings rated by severity and prioritised recommendations.

✓

Opinion on fitness for use

✓

Findings rated by severity

✓

Remediation plan with deadlines

✓

Follow-up on implementation

References:

ECB guide to internal models

IFRS 9

AI Act

Model risk management best practice

To preserve independence, we do not validate models developed by the same Gryphon team.

Methodology

How a validation works

Phase 1

· 1 week

Scope and materiality

We define the perimeter, depth and criteria, proportionate to the model’s risk.

Phase 2

· 2–4 weeks

Review and replication

Review of documentation and data, and independent replication of results.

Phase 3

· 2–3 weeks

Testing

Performance, stability, sensitivity, benchmarking and, for AI, explainability and bias.

Phase 4

· 1–2 weeks

Report

Opinion, findings and remediation plan, presented to the responsible bodies.

Indicative timelines for a model of medium complexity.

Frequently asked questions

Before you start

Can Gryphon develop and validate the same model?

+

What types of models do you validate?

+

Our models come from a vendor. Do they still need validation?

+

When were your models last validated?

Talk to us to scope an independent validation, proportionate to the materiality of your models.

Book a diagnostic →

© 2026 Gryphon Consulting

Lisbon, Portugal · Europe