What Changed

Chai Discovery deployed Chai-3, a next generation antibody design model that the company describes as roughly doubling the success rate of Chai-2 and binding more tightly to harder targets. Date recorded here is the date of the trade coverage confirming deployment; the company has not published a precise release date for Chai-3, and this record does not assert one.

Our read

For an R&D buyer the question is whether a model generation actually changes the design cycle or only the benchmark. The reported doubling is a vendor claim measured against the company's own prior model, not an independent comparison against competing methods, and no candidate designed by these models has published clinical results. What is verifiable is that named pharmaceutical organizations have licensed the platform into production R&D, including at least one custom model trained on a customer's proprietary data. That structure is the thing to scrutinize in procurement: a custom model trained on internal discovery data creates a different dependency, and different exit economics, than licensing a general model.

Nabla appears to have withdrawn public per month pricing in favor of a demo and contract sales motion, particularly for organizations. The company historically published a free tier of roughly 30 consultations a month and a Pro plan reported near $119 per month. Third party listings now conflict on tier names and rates. Recorded as Partially Verified: the shift is consistently reported by third parties but was not confirmed in vendor published materials, and the date reflects the verification pass rather than an announcement.

Our read

This is a change in the buyer's favor being reversed, and it is worth noticing because the ambient scribe category is moving in two directions at once. Heidi restructured pricing upward in February 2026 but kept published tiers; Nabla appears to have withdrawn public pricing entirely. For a small practice or an individual clinician, that removes the ability to budget without a sales conversation and eliminates a self serve evaluation path that previously existed. For a health system it changes little, since enterprise contracts were always negotiated. The more consequential items to settle in that sales conversation are not the rate: confirm whether a Business Associate Agreement is included at your tier, and whether the default use of session data for model training can be contractually disabled.

HeartSciences released MyoVista Insights version 1.3 with an AI-ECG Algorithm Marketplace, making Bunkerhill Health's FDA cleared ECG-EF algorithm for detecting reduced left ventricular ejection fraction available through the platform. Both companies are indexed here, making this a distribution relationship between two vendors in this index: Bunkerhill developed and cleared the algorithm, HeartSciences distributes it.

Our read

This is a distribution event rather than a clearance, and the distinction matters. The cleared algorithm belongs to Bunkerhill; HeartSciences supplies the platform that delivers it. For a community hospital, critical access hospital, or primary care practice, that is the substantive change: a cleared AI-ECG model detecting reduced ejection fraction becomes reachable without a direct enterprise agreement with the algorithm developer, in a setting where echocardiography is often not readily available. Buyers evaluating MyoVista Insights should be clear that the AI they gain access to today is Bunkerhill's, since HeartSciences' own algorithms for impaired cardiac relaxation and low ejection fraction remain in development and unsubmitted or in pre submission. Assistive AI-ECG assessments are reimbursable under the Hospital Outpatient Prospective Payment System at APC 5734.

Subtle Medical received FDA clearance for SubtleHD (PET), a next generation AI PET image enhancement product for PET/CT and PET/MR systems. The company states it supports all FDA approved radiotracers, enables up to 75 percent faster PET imaging on existing scanners, improves SUVmax quantitation accuracy, allows radiologist adjustable denoising levels, and incorporates anatomical CT data into reconstruction. It extends prior clearances covering SubtlePET and SubtleHD (MR).

Our read

PET scanner capacity is a hard constraint for most imaging programs, and acquisition acceleration is one of the few levers that adds throughput without capital equipment. The material detail for buyers is radiotracer compatibility: support across all FDA approved radiotracers, rather than 18F-FDG alone, is what makes this usable in theranostics and advanced molecular imaging programs where the older generation was limited. Two caveats worth carrying into procurement: the acceleration figure is a vendor claim and should be validated against your own protocols and reader acceptance, and adjustable denoising means image appearance becomes a configurable parameter, which needs a documented standard rather than per radiologist preference drift.

CMS established a national billing code and associated payment under the Hospital Outpatient Prospective Payment System for algorithmic analysis of coronary artery calcium and aortic valve calcium on chest CT, effective April 1, 2026. The company separately received FDA clearance for Bunkerhill Contrast CAC and Bunkerhill Contrast AVC, which it states are the first AI algorithms cleared to detect and quantify coronary artery calcium and aortic valve calcium on contrast enhanced, routine non gated chest CT, extending prior clearances that covered non contrast chest CT.

Our read

Two things changed here and the second matters more. FDA clearance on contrast enhanced non gated chest CT means the algorithm can run opportunistically on scans a health system is already acquiring for other reasons, rather than requiring a dedicated gated cardiac study. That converts an existing imaging volume into a screening surface without new acquisition cost. The CMS billing code is the rarer event: most healthcare AI carries no reimbursement pathway at all and must be justified purely on internal ROI, so a national OPPS payment moves this from a cost center argument to a billable service line. Buyers evaluating this platform should separate the two economic questions, since the reimbursement applies to specific algorithmic analysis and not to the cost of licensing the platform itself.

Caris added two Caris AI Insights signatures assessing brain metastasis risk in breast and lung cancer, bringing its proprietary signature count to seven. The signatures were trained on 12,994 NSCLC cases and 3,371 breast cancer cases with matched survival outcomes, and generate a personalized predictive score from a patient's whole exome and whole transcriptome data, visualized as Kaplan Meier curves. They are delivered in the Caris Molecular Tumor Board Report, a research use only report available at no additional cost when ordering MI Cancer Seek.

Our read

The signature itself matters less than what it reveals about how to read this vendor. Caris publishes training cohort sizes per signature, which is more disclosure than most AI diagnostics offer and lets a clinician judge whether their patient population is represented. But the signatures ship inside a research use only report, while the underlying MI Cancer Seek assay carries FDA approved companion diagnostic indications. Those are different regulatory objects arriving in the same envelope at no extra cost, and the distinction governs how a result can be used in a treatment decision. Confirm with your pathology and legal teams which outputs are approved for clinical use before signatures reach a tumor board.

Elation Health

Mar 6, 2026

MediumPartially Verified

Elation Health launched AI Fast Lane, applying Smart Coding to primary care billing. The system reads pre visit details, the patient problem list, active medications, and the visit note (often drafted by its Note Assist ambient scribe) and suggests diagnosis, procedure, and drug codes, shifting the biller's task from building a claim to confirming a suggestion. Claims that meet a defined confidence threshold bypass the manual review queue and are submitted automatically.

Our read

For an independent practice already on Elation, this is the case for consolidating rather than buying a separate coding tool, and the AI carries no additional line item. The provision that deserves scrutiny in procurement is the touchless lane itself. Claims above a confidence threshold submit without human review, which is where the efficiency comes from and also where the compliance exposure sits. Ask who sets the threshold, whether it is auditable, what the error rate is on auto submitted claims, and who is accountable for a miscoded claim that was never seen by a person. Bundled AI at no extra cost is a genuine economic advantage over a specialist tool, but only where the practice was going to run this EHR regardless.

Heidi Health

Feb 1, 2026

HighPartially Verified

Heidi restructured its pricing in February 2026 and moved upmarket. The main paid tier, previously reported near $90 a month, became Clinician at $150 per user per month billed annually, with an Evidence Plus tier reported near $40, a Practice tier adding EHR integrations, and quote based Enterprise plans with SSO and custom hosting. The free tier remains, with unlimited transcription and 10 Pro Actions a month. Sources conflict on some tier amounts.

Our read

This is the single most consequential change for anyone already using or evaluating Heidi. The free and individual tiers remain genuinely usable for trialling the product on real clinic days, which is rare in this category. But two gates moved upmarket: push to chart EHR integration and an organization wide Business Associate Agreement are both reported as available only from the Practice tier upward. For a US organization, that means the entry tiers are not a viable production configuration for PHI handling regardless of how well the scribe performs, and the evaluation should be run against the tier you would actually buy.

Aidoc

Jan 21, 2026

HighVerified

The FDA cleared Aidoc's comprehensive abdomen CT triage solution, powered by its self developed CARE foundation model, bringing 11 newly cleared indications together with three previously cleared into a single 14 indication workflow. The company states this is the first FDA clearance of a double digit set of acute indications powered by one foundation model. In the FDA reviewed pivotal study the new indications reported a mean sensitivity of 97 percent and mean specificity of 98 percent, and the company reports roughly an order of magnitude reduction in false alerts compared with leading single condition tools.

Our read

The operationally decisive number here is not the sensitivity, it is the false alert rate. Triage AI fails in practice when radiologists learn to ignore it, and single condition tools deployed side by side compound that problem because each one alerts independently. A reported order of magnitude reduction in false alerts, if it holds in a buyer's own case mix, is what determines whether the tool survives past month three. Two things to establish in evaluation: that the cleared indications match the acute findings actually driving delays in your emergency department, and how the platform governs models it did not build, since aiOS also hosts third party AI and the governance layer is doing work across all of them.

ArteraAI

Dec 18, 2025

HighVerified

The ArteraAI Prostate Test was included in the NCCN Clinical Practice Guidelines in Oncology for Prostate Cancer, making it the first AI enabled risk stratification tool recommended in those guidelines. It carries a Level 1B evidence rating under Simon Criteria and a Category 2A recommendation. This followed FDA De Novo marketing authorization in August 2025 and CMS coverage effective January 2024.

Our read

Guideline inclusion is a different class of validation from a vendor claim or even a regulatory clearance, because an independent panel evaluates the underlying evidence and decides whether clinicians should consider the tool at all. For a center weighing AI diagnostics, this is the clearest available answer to the question of whether a test has been assessed by people with no commercial interest in the answer. The specific standing matters: Level 1B evidence under Simon Criteria and a Category 2A recommendation indicating uniform NCCN consensus that the test is a useful option. Note also what the guideline language constrains, since it positions the test as used in addition to NCCN risk category, patient characteristics, and patient preference, which is an externally imposed limit on how far the output may be relied upon.

Pearl

Dec 18, 2025

MediumVerified

Pearl received FDA clearance extending Second Opinion to panoramic radiographs, covering detection of suspected caries, periapical radiolucencies, and impacted third molars. The clearance was supported by a standalone performance study and a fully crossed multi reader, multi case study, with the company reporting stable performance across gender, geography, and imaging device subgroups. With this clearance the platform spans bitewing, periapical, panoramic, and CBCT imaging.

Our read

The clinically interesting part is the study design, not the clearance itself. A fully crossed multi reader, multi case study is the appropriate design for a reader assistance claim, because it measures whether clinicians perform better with the tool rather than whether the model performs well alone, and relatively few AI vendors in any specialty run one. The reported stability of performance across gender, geography, and imaging device subgroups is also a genuine fairness disclosure, which remains the emptiest axis across this index. Buyers should still confirm that the imaging devices in their own practice are represented in the validation population.

Cohere Health

Sep 1, 2025

MediumPartially Verified

Cohere Health acquired ZignaAI in September 2025 and launched the Cohere Payment Integrity Suite, extending its clinical AI from pre service prior authorization into post service claims and coding validation. The company subsequently launched specialized audit agents for high cost conditions including sepsis, reporting a 58 percent findings rate, and states the combined pre and post service view achieves up to 9x return on investment.

Our read

Two things matter here beyond the acquisition itself. First, scope creep across the claim lifecycle: a platform that decides authorization before care and then audits payment after care sits on both sides of the same clinical event, which is efficient for the plan and worth examining for providers, since the evidence surfaced at authorization is explicitly reused to validate the claim. Second, the audit agents carry a stated 58 percent findings rate on conditions like sepsis. A findings rate is a measure of how often the model flags something, not how often it is right, and sepsis coding is among the most contested areas in payment integrity. Plans should establish the confirmation rate on appeal, not just the findings rate, before treating that number as savings.

Tempus

Mar 1, 2025

MediumVerified

Tempus acquired Deep 6 AI in March 2025, adding an AI precision research platform that mines structured and unstructured EMR data to match patients to clinical trials and generate real world evidence. Deep 6 continues to operate under its own name and reports real time EMR feeds across more than 30 health systems and an ecosystem of over 1,000 research facilities. This index records the capability inside the Tempus vendor entry rather than as a separate record, consistent with how other acquired product lines are handled.

Our read

For a health system already running Tempus Next or Hub, this quietly widens what the relationship covers: the same vendor now sits on clinical decision support, prior authorization, and trial recruitment, all reading the same EMR. That consolidation is convenient and worth scrutinizing, since the data access granted for one purpose now serves several. The technically interesting claim is about unstructured data: Deep 6 reports that 92 percent of trial inclusion and exclusion criteria benefit from unstructured sources and that 15 to 20 percent of eligible patients are found through free text alone, which is the actual argument for NLP based matching over coded data queries. Two things to establish: that data ownership remains with the health system rather than transferring with the platform, and that matched cohorts are treated as decision support requiring validation, which the company itself states.