Diagnostics & imaging AI · Breast cancer screening AI
Vara
Vara is a Berlin company whose AI sorts screening mammograms into normal, suspicious and uncertain cases for radiologists, used across Germany's national screening program.
Strong fit
- Breast screening programs in Europe that read large volumes of normal exams
- Screening units that want AI to set aside clearly normal mammograms
- Programs that want screening workflow, reporting and AI on one platform
Weak fit
- US imaging centers, since Vara is not sold in the US
- Radiology groups that want one AI vendor for many body areas
- Buyers who need a posted price
Bottom line
Vara is a strong choice for European breast screening programs that want AI to triage normal mammograms and catch cancers readers might miss. It suits high-volume units that want evidence from a large national study. US centers cannot buy it, and every buyer will need to negotiate a price.
Score breakdown
Weights: Outcomes 35% · Product 30% · Implementation 20% · Pricing clarity 15%.
Jonas Muff and Stefan Bunk founded Vara in Berlin in 2018 after meeting at Merantix, a Berlin venture studio for AI companies. Muff is chief executive and Bunk is chief technology officer, and the company has an office in India.
Vara reads screening mammograms and sorts them. Exams it judges normal with high confidence are labeled normal for the radiologist, and exams it finds highly suspicious trigger a safety net that points the reader to the area of concern if they mark the exam negative. Cases in between go to the radiologists with no AI label.
The PRAIM study, published in Nature Medicine, followed 463,094 women in Germany's screening program. Units using Vara found 6.7 cancers per 1,000 women against 5.7 without it, a 17.6% rise, and recalled slightly fewer women. Vara says it runs in more than half of Germany's screening units and reads more than 250,000 exams a month.
In September 2026 Vara received a CE mark, Class IIb, for autonomous triage of normal exams, with a monitoring service that tracks its performance after launch. The platform reads 2D mammograms and 3D tomosynthesis and includes worklists, structured reports and follow-up tracking for screening units.
Vara is not cleared for sale in the US, so it fits European, Latin American and Asian screening programs. It has about 50 staff and has raised about $34 million.
ScreenPoint Medical fits programs that want another widely studied mammography AI, and Therapixel fits US and French centers that want FDA-cleared mammography AI. Koios Medical fits teams that want AI for breast ultrasound.
Competitor landscape
| Vendor | Overall | Ease of implementation |
|---|---|---|
| ScreenPoint Medical | 7.9 | 7.6 |
| Vara | 7.7 | 7.3 |
| Therapixel | 7.5 | 7.2 |
| Koios Medical | 6.7 | 6.3 |
Pricing
| Item | Detail |
|---|---|
| Model | Vara sells its screening platform by contract to screening units and health systems. |
| What usually drives cost | For Vara, cost depends on exam volume, whether you use autonomous normal triage, and how much of its workflow software you take on. |
| What to ask before you buy | Ask Vara for its price per exam, how its monitoring service tracks performance at your site, and which of your mammography machines it supports. |
| Published pricing | Vara does not publish a price list. |
Leadership
Methodology
| Weight | Factor | What it measures |
|---|---|---|
| 35% | Customer outcomes | Whether buyers get measurable operational or clinical-workflow results after go-live |
| 30% | Product | Capability depth, reliability, and fit for the job the category buys |
| 20% | Implementation | How hard it is to stand up, integrate, train, and stabilize |
| 15% | Pricing clarity | Whether a buyer can model total cost without a mystery quote |
| Label | Meaning |
|---|---|
| Highly recommend | Strong outcomes and product with manageable caveats |
| Recommend | Solid fit for the right buyer; know the tradeoffs |
| Conditional | Only with a specific use case or heavy caveats |
| Not recommended | Avoid for most buyers in this category |
Read our full methodology for how we weight scores and assign recommend labels.