For TPAs and insurance carriers

The AI-native claims system of record.

We help TPAs and carriers close claims faster. Our agents handle the procedural work, your adjusters approve every consequential action, and routine tasks run on models you own.

OpenAI Select PartnerOpenAI
Recognized across the industry
NAMICInsurtech InsightsThe CLM — Claims and Litigation Management Alliance

01Why now

Unlike wine, claims don't age well. The longer they stay open, the more they cost.

We build the system of record with AI inside it, one workflow at a time, so each claim moves without waiting for procedural work.

Agents

Procedural work, handled

Agents handle intake, dispatch, file review and correspondence.

Adjusters

People make the decisions

Adjusters approve every consequential action. Their corrections become new training data.

Models

Routine tasks on models you own

Fine-tuned open models do routine claims work for a small part of the frontier price.

Governance

Ready for audit

Each model ships with its eval suite, data lineage and change log, so your compliance team can review it.

02How we work

Evaluate. Train. Deploy. Improve.

We start from one workflow your claims team runs today. Each step has a measurable output before the next one starts.

1

Evaluate

We turn 10 to 20 of your real claim tasks into an eval, then run frontier and open models against it on your side.

Output · cost and quality per model
2

Train

We post-train an open model with fine-tuning and reinforcement learning, using your data, graders and adjuster feedback.

Output · a model that beats the baseline
3

Deploy

We serve the model in your cloud, inside your workflow. Hard cases go to a frontier model. Your team approves every consequential action.

Output · a model in production
4

Improve

Production traces and adjuster corrections become new training data. Each release must pass the eval before it ships.

Output · a model that gets better with use

OPEN WEIGHTS · YOUR CLOUD OR OURS · MODEL-NEUTRAL · YOU KEEP EVERYTHING WE BUILD

03Proof

Our claims model beat every frontier model we tested. At about a ninth of the cost.

We fine-tuned a 27B open model on workers' compensation claims work and tested it on 496 held-out tasks against 14 frontier configurations.

Read the study →
Cost per 1,000 tasks
Hypermodel 35B MoE, fine-tuned$1.62
Hypermodel 27B, fine-tuned$5.71
GPT-6 Astra · medium$50.90

04Who we serve

For the teams that handle the claim.

TPAs

Claims handling is your product

Cost per file sets your margin, and adjusters still do procedural work by hand.

  • Intake and dispatch
  • File review
  • Correspondence
  • Models you own, trained on your files

One workflow at a time · audit trail · your weights

Carriers

Close claims faster

Open claims cost more every day. We take the procedural work off your adjusters' desks.

  • Shorter cycle time
  • Adjusters approve every consequential action
  • Measured cost and quality per task
  • Deployment in your cloud

VPC deployment · audit trail · model risk documentation

05Why Hypermodel

Built inside claims. Led by a regulated-industry operator.

Licensed

Built by a licensed adjuster

Our founder holds a claims adjuster license and a California broker license.

Regulated by default

Stripe and Wise experience

Our founder built products at Stripe and Wise, under payments regulation. We know what an audit asks for.

Model-neutral

OpenAI Select Partner

We use frontier models where they earn their cost, and open models you own where they do not.

Yours to keep

No lock-in

You keep the weights, the eval suite, the datasets and the training recipe.

06Get started

Start with a free model audit.

Bring 10 to 20 of your own claim tasks. We run them on your side and show you where a model you own wins on cost and quality.

  • 01Cost and quality per task for each model
  • 02Runs in your environment. Your data stays with you
  • 03Yours to keep, whether or not we work together