Nabla and Suki both sell clinical documentation technology to healthcare organizations, but their public documentation describes different integration surfaces. Nabla positions its Core API as a set of transcription, note generation, normalization, and coding endpoints that developers call directly, with dated API versions controlling when changes take effect [1][2]. Suki describes a partner platform built around Ambient Clinical Intelligence, with SDKs, REST APIs, and webhooks that support ambient notes, form filling, dictation, and patient summaries [10][9]. The practical comparison is not a single feature checklist but a set of decision points: whether a team wants a versioned server API, hosted UI components, or a mix of both; how each vendor structures coding output; and what each documentation set leaves unspecified. This article examines those documented differences without inventing test results, prices, or clinical conclusions.

Integration posture: versioned API versus partner SDK platform

Nabla's Core API documentation describes a server-side integration model. Developers can transcribe encounters, generate clinical notes, extract FHIR-normalized data, and request coding through endpoints such as the WebSocket Transcribe endpoint for real-time audio and REST endpoints for recorded files [1]. The changelog states that updates are released under new-dated versions, so changes do not affect an integration unless the developer explicitly opts into a newer version [2]. This versioning model gives integrators a predictable upgrade path, but it also means that newer capabilities, such as the ICD-10-CM 2027 code set, require an explicit version move to take effect [2].

Suki's documentation describes a different posture. The Suki for Partners platform includes APIs, SDKs, and infrastructure for building AI-powered clinical workflows into EHRs, telehealth platforms, and revenue cycle management solutions [10]. Developers can choose between client SDKs and backend REST APIs, with the Web SDK shipping hosted UI components and the Headless Web SDK giving more layout control for React 18+ applications [10]. Suki also lists product availability explicitly: Ambient APIs, Web SDK, Mobile SDK, and Form filling APIs are generally available, while Headless Web SDK, Dictation SDK, Form filling SDK, Patient Summary API, and E/M Coding APIs are in beta [9]. This distinction matters for teams planning production rollouts.

Coding and structured output differences

Nabla's normalization endpoint extracts ICD-10 and LOINC codes from a Nabla-generated clinical note in a FHIR-compliant structure [4]. The documentation states that the endpoint identifies past medical history, family medical history, clinical assessments, and vital signs, representing each piece of medical information as an array of FHIR conditions or observations [4]. A changelog entry notes that a new ICD-10 normalization engine, available on version 2026-07-01 and later, suggests more specific codes and covers the full ICD-10-CM code set, with a median processing time of about five seconds per note [2]. Nabla also documents asynchronous normalized data generation endpoints so that normalization does not block a workflow [2].

Suki's coding documentation describes E/M Coding as a separate capability that recommends Evaluation and Management codes for a clinical visit [9]. The documentation states that E/M Coding supports revenue cycle management workflows by returning coding recommendations after clinical documentation is available [9]. Suki also describes Problem-Based Charting as a way to organize notes by patient problems and return coded diagnoses for EHR workflows [12]. The two vendors therefore document coding differently: Nabla emphasizes ICD-10 and LOINC extraction from generated notes, while Suki's public feature overview emphasizes E/M code recommendations and problem-based charting. Neither documentation set provides comparative accuracy data.

A suggested evaluation workflow for a healthtech team

A team building an EHR integration could start by mapping its required workflow to each vendor's documented surface. If the team needs a versioned server API with explicit control over when changes take effect, Nabla's dated API versions and changelog provide a clear mechanism [2]. If the team wants hosted UI components for a React or JavaScript application, Suki's Web SDK is documented as generally available, while its Headless Web SDK is in beta [9][10]. The team could then test a single encounter flow: create a session, stream audio, retrieve the generated note, and inspect the structured output. For Nabla, the output would include FHIR-compliant conditions and observations from the normalization endpoint [4]. For Suki, the output would include LOINC-coded sections and, if enabled, structured diagnoses and orders [12]. This suggested workflow is a planning exercise, not a substitute for a controlled pilot with real clinical audio and clinician review.

EHR integration and workflow context

Nabla's EHR integration page describes an AI layer that runs inside the EHR clinicians already use, with documentation, dictation, and coding flowing into the chart [3]. The page lists Epic, Oracle, athenahealth, NextGen, and Greenway as EHRs, and describes Nabla Connect as a way for EHR vendors to embed ambient AI through a single iframe [3]. The Connect deployment process involves configuring authentication, creating an encounter via API, launching Nabla within the EHR, and receiving the generated clinical note [5]. Nabla's Epic page further distinguishes between AVR, which records natively inside Haiku, and a standard integration that works on mobile, desktop, and telehealth [7].

Suki's ambient documentation guide describes a similar context flow: seed session context before or during capture, stream audio while the visit runs, and push approved results back into clinical systems [12]. Common context inputs include patient demographics, provider identity and specialty, visit type, and encounter identifiers [12]. Suki also documents interoperability across supported products, allowing a note to start in one product and continue in another [9]. The key difference is that Nabla's documentation emphasizes a single iframe embed for EHR vendors, while Suki's documentation emphasizes a choice between hosted UI components and headless control, with interoperability across its own product set.

What the documentation does not establish

Neither vendor's public documentation provides a direct comparison of note quality, transcription accuracy, or coding precision. Nabla's marketing pages cite deployment statistics such as 89% same-day note completion across Nabla deployments and a 47% reduction in clinician burnout at CHLA, but these are vendor-attributed figures without independent verification in the public documents [3][5]. Suki's documentation mentions a patent portfolio of 50+ patents and describes itself as a leader in Ambient Clinical Intelligence, but this is a vendor claim rather than a measurable product difference [10]. Buyers should treat these statements as marketing context, not comparative evidence.

Several practical questions remain unanswered by the supplied documentation. Neither public documentation specifies pricing, contract terms, or total cost of ownership. Neither documents a head-to-head evaluation of the two products. Nabla's documentation states that its E/M code generation endpoints are available in the U.S. product release only, and Suki's documentation does not specify which medical specialties its supported specialty list covers [2][12]. Teams evaluating both vendors will need to request additional information directly, particularly around data retention defaults, model training policies, and the specific EHR write-back fields each platform supports.