BioMap

xTrimo is the company's reason for being. A single large protein model is trained on biological data and a set of task models draw on it to predict how proteins behave and to design and optimize biologics. The company's whole pitch is that this map of protein biology is what finds targets and shapes drugs. Remove the model and there is no business left, which is how BioMap has described and funded itself from the start.

xTrimo predicts how proteins behave and designs biologics, and a partner leaning on those outputs would want to know how they are checked: where a scientist reviews or overrides what the model returns, what happens when a prediction fails downstream, and which steps run with no person involved. BioMap conveys that it runs a careful process but does not set out any of these as a defined control, so the supervision stays a matter of trust rather than a mechanism a partner can point to.

The model is named and placed in a clear class. BioMap calls xTrimo a protein focused large language model, the largest of its kind by its own account, and describes an architecture in which one foundation model trained on unlabeled data feeds a family of downstream task models. The kind of model and how it is structured are both identified.

The quantities are not published here. No parameter count, training data size or cutoff is stated on the company's pages, and no version or update practice ties a given result to a dated model, so the shape of the system is clear while its specifics are not.

The platform is presented as BioMap's own, a foundation model it trains and runs, and no outside model provider is named as sitting between data and a design. Beyond that nothing is established: the super scale computing the company refers to is not attributed to a named provider, no third party components or licenses are listed, there is no subprocessor list, and nothing describes what governs a partner's data once it enters the system.

The evidence is a major partner rather than a clinical result. Sanofi signed a collaboration in October 2023 to co develop AI modules with BioMap, paying $10 million upfront against a deal worth more than $1 billion, which is a large pharmaceutical company committing real money to the platform.

What is missing is the rest: no program from the platform is named in a clinical trial, and no peer reviewed evaluation of how well xTrimo performs on a drug design task is cited, so the validation a reader can check is the partnership, not the performance behind it.

Because xTrimo is a single foundation model that many projects draw on, the data question has a sharp edge: a partner's proprietary data goes into the same system that serves everyone else. Whether that data is retained, used to train the shared model, or kept apart from other collaborations is the thing a partner needs answered, and BioMap does not answer it. The inputs are biological and protein data rather than patient records, so no protected health information is involved, but the stewardship of proprietary partner data is left open.

BioMap works with biological and protein data rather than patient records, so United States health privacy rules do not bear on what it handles. No business associate position and no service level privacy document appear on the site, so a partner has nothing published to point to on this question.

BioMap holds proprietary data from Sanofi, a large pharmaceutical partner, which is the kind of relationship where a security team expects an independent attestation on file: a SOC 2 report, an ISO 27001 or HITRUST certification, or a trust portal to review. None of these is named, and the company publishes no security page, so there is no outside account of how the environment holding that data is secured.

No device claim is made, and the platform is scoped to match. BioMap works at the discovery stage, finding targets and designing biologics, so any regulatory position belongs to the molecules and the partners that develop them rather than to xTrimo. There is no clearance to hold and none is claimed.

xTrimo is a large protein language model, the kind of system with a research lineage that often carries published benchmarks. A buyer wanting to judge how reliably it performs, how its behavior is tested, or what responsible AI rules govern it finds none of that on BioMap's pages. Whatever evaluation sits in the model's research history is not carried forward here as a measure of production reliability, so the company's own pages give a buyer nothing to gauge model governance by.

A wrong prediction costs a partner wet lab time and budget, so the natural questions are how often the platform misses and what the company stands behind. BioMap publishes no success or error rate carrying a method and a denominator, which leaves no figure to plan against, and no guarantee, indemnity or correction route travels with an output. Those terms live inside the private collaboration agreement rather than on the public site.

The platform is a foundation model with a family of task models on top, not an application a buyer connects to a stack, so the absence of a clinical record surface is what a discovery system would look like. Nor is there a research side integration to point to: no electronic lab notebook, laboratory information management system or research data platform is named as connected, and no connector is offered to a partner who would want outputs to land in their own systems.

The platform runs for BioMap and its partners rather than being deployed to anyone, so there is no tenancy or hosting model to evaluate. Location is implied by the company's presence in California and China rather than committed, and nothing published names where partner data rests or how one collaboration's data is separated from another's on the shared foundation model.

BioMap earns by co developing with partners rather than selling access, so no price is quoted. The figures that are public come from the Sanofi deal, $10 million upfront against more than $1 billion in potential payments and milestones, which are contract terms rather than a price a buyer could pay, so the business model is visible while the cost of engaging the platform is not.

Coverage is described broadly and without the specifics that would pin it down. BioMap presents xTrimo as a general map of protein biology used to find targets and design biologics across many tasks, which is a wide claim, but no particular disease area, target or validated program is named to show where that breadth has been demonstrated. The reach is asserted at the level of the whole field rather than named target by target.