Nabla
The only ambient scribe that beat a control group in a randomised trial, and the only one with a real public API — but it stopped publishing its price.
PRICING
| Free | Free | ~30 consultations/month with EHR integration; unlimited free for interns and residents (third-party figure) |
| Pro | ~$119/mo | per clinician, unlimited consultations, no per-note metering (third-party figure; vendor publishes nothing) |
| Enterprise / Nabla Connect | Custom | sales-led: deeper EHR integration, tailored models, SSO/SAML, admin controls, BAA |
Nabla used to publish a price and no longer does — the pricing page is a 404 as of this audit. Every number above is third-party. Treat them as a range to open a conversation with, not a quote.
checked 2026-08-12 · vendor pricing page
Element scores
Strengths
Nabla has the one thing nobody else in this category has: a randomised controlled trial where it beat the control group. The UCLA trial published in NEJM AI put 238 physicians across 14 specialties and roughly 72,000 encounters into three arms — Nabla, Microsoft DAX, and usual care. Nabla cut time-in-note by 9.5% versus control (95% CI -17.2% to -1.8%, P=0.02). DAX did not reach significance. Both arms showed burnout improvement. That is a harder claim than any competitor's case-study deck. The second differentiator is the developer surface. The Nabla Core API is a real product — WebSocket and REST transcription, note generation, Magic Edit, FHIR normalisation with ICD-10 and LOINC codes, dot phrases, a custom dictionary, OpenAPI specs, a Postman collection, a public GitHub sample app and 19 dated API versions. Nabla Connect drops the whole assistant into a partner EHR in about three days. In a category where the incumbent has no public API at all, that is the difference between building on a vendor and renting one. Scale and transparency back it up: 190+ organisations, 100,000 clinicians, 40M encounters a year, 35+ languages, a published CHAI Applied Model Card, and an external auditing firm running weekly three-layer audits. At the University of Iowa, burnout fell 26% and held at two years.
Honest dings
Nabla stopped publishing its price. The pricing page is a 404 today, which means a solo clinician who could once read a $119 number and decide now has to guess or call sales — and third-party estimates for the Pro tier range from $119 to $239 depending on who you read. For a company whose free tier is its best land-and-expand asset, that is a self-inflicted wound. On the periodic table, the gaps are structural rather than sloppy: no customer-facing evaluation suite, no model ladder or cost tier, no MCP or A2A, and no fine-tuning you can run yourself. Reviewers consistently report EHR write-back is lighter than Abridge or DAX on some systems, closer to a structured paste than native field-level writing, and the coding features are newer than the incumbents' revenue-cycle machinery. There is no human-QA review layer — correction is entirely on the clinician, and the same RCT that produced the good headline also recorded clinically significant inaccuracies happening 'occasionally' (Likert 2.8) and one mild adverse event. The AMI Labs world-model partnership is a 2027 story, not something you can use in August 2026.
Sources (18) — every claim traceable
Every audit lists the research it rests on — transparency and traceability are the product. Tools evolve: each audit is a snapshot of its audit date, and re-audits supersede older versions (kept below for reference).
- pubmed.ncbi.nlm.nih.gov/41497288 — Primary independent efficacy evidence. 'Ambient AI Scribes in Clinical Practice: A Randomized Trial', NEJM AI 2025, DOI 10.1056/AIoa2501000, ClinicalTrials.gov NCT06792890 — 238 physicians, 14 specialties, three arms (DAX / Nabla / usual care), Nabla -9.5% time-in-note vs control (P=0.02), DAX not significant; clinically significant inaccuracies 'occasionally' (Likert 2.8), one grade 1 adverse event (accessed 2026-08-12)
- uclahealth.org/news/release/ucla-study-finds-ai-scr… — Independent institutional write-up of the same RCT: ~72,000 encounters, Nabla note time 4m30s to 3m49s vs control 4m22s to 4m04s, ~7% burnout-score improvement in both AI arms (accessed 2026-08-12)
- docs.nabla.com/guides/intro — Official Nabla Core API documentation: WebSocket + REST + async transcription, clinical note generation, Magic Edit natural-language editing, patient instructions, FHIR normalisation with ICD-10 and LOINC, dot phrases, custom dictionary, feedback reporting, OpenAPI specs, Postman collection, GitHub sample app, 19 dated API versions (accessed 2026-08-12)
- nabla.com/connect — Official Nabla Connect page: single-API embed for EHR platforms, iframe session model, ambient AI live in 3 days, 35+ languages, multi-speaker separation, smart code suggestions, SOC 2 Type 2 / ISO 27001 / GDPR / HIPAA (accessed 2026-08-12)
- medonesystems.com/docs/MedOne-IRM-Requirements-Nabl… — Third-party governance review (MedOne Systems, completed 11/03/2026) mapping Nabla's Nov 2025 v2 model card: 95% average transcription accuracy, PDQI 48/50, validated across 3,442 clinicians and 303,266 encounters in 6 specialties, weekly audits averaging 5% risk findings, external auditing firm with three verification layers, human-in-the-loop required, client data not used for training by default (accessed 2026-08-12)
- www-assets.nabla.com/docs/Nabla%20whitepaper_Raisin… — Official evaluation whitepaper: 7,000 hours of scribe-transcribed conversations, 50,000 production situations with a 400-case iteration subset, LLM-judged hallucination and translocation tests, blinded expert side-by-side before deployment, progressive rollout with edit-depth tracking (accessed 2026-08-12)
- nabla.com/press-release/nabla-appoints-brian-mannin… — Official 21/07/2026: 190+ healthcare organizations, 100,000 clinicians, 40M patient encounters annually; CVS Health, M Health Fairview, UCLA Health, University of Iowa Health Care; Brian Manning appointed CEO, Alex LeBrun to Executive Chairman/Chief AI Officer; exclusive AMI Labs partnership (accessed 2026-08-12)
- statnews.com/2026/07/21/ai-scribe-nabla-new-ceo-pla… — Independent (STAT) 21/07/2026: Paris-based, new CEO concedes 'we really haven't built our brand in the United States' and points go-to-market acceleration at 2027 (accessed 2026-08-12)
- trust.nabla.com — Official trust centre: CHAI version 2 AI Model Card published, dedicated AI Governance team, stated position that Nabla is not high-risk under the EU AI Act but follows published model transparency standards (accessed 2026-08-12)
- registry.chai.org/applied-model-card — Independent confirmation that Nabla is listed in the CHAI Applied Model Card registry alongside Kaiser Permanente, Mount Sinai, Stanford Medicine and Providence (accessed 2026-08-12)
- nabla.com/pricing — Checked directly on 12/08/2026: returns HTTP 404, 'Sorry, the page you're looking for isn't available'. Nabla publishes no list price (accessed 2026-08-12)
- litmustools.com/review/nabla — Third-party review 02/07/2026 (treat pricing as estimate): free plan ~30 consultations/month with EHR integration, unlimited free for interns and residents, Pro $119/month per clinician month-to-month; notes lighter EHR write-back than DAX/Abridge, newer coding features, no human-QA review layer (accessed 2026-08-12)
- aiagentsquare.com/agents/nabla-ai — Independent review 26/10/2025 giving a conflicting third-party price band (~$119 Starter, ~$239 Pro) and confirming SMART on FHIR integration with Epic, athenahealth, Cerner and Meditech — cited here as evidence that unpublished pricing produces divergent estimates (accessed 2026-08-12)
- nabla.com/case-studies/iowa-case — Vendor case study (treat as vendor-reported): University of Iowa Health Care, 2,200 clinicians, 26% burnout reduction sustained at two years, 100% of pilot clinicians continued post-rollout, 30% system-wide adoption across 25+ specialties in five months (accessed 2026-08-12)
- ama-assn.org/practice-management/digital-health/ai-… — Independent corroboration (AMA, 24/04/2025): ~50% regular outpatient adoption at University of Iowa, burnout 69% to ~42% in the trial group, documentation 'pajama time' reduced by hours (accessed 2026-08-12)
- beckershospitalreview.com/healthcare-information-te… — Independent (Becker's, 08/07/2025): 220,000 encounters since launch (~a third of clinical volume), average 2.6 hours/week saved on after-hours documentation, category pricing $100-$600 per provider per month (accessed 2026-08-12)
- pmc.ncbi.nlm.nih.gov/articles/PMC12973079 — Independent narrative review 19/01/2026 placing Nabla among DAX, Abridge, ChatGPT-4 and TORTUS; documents the Misurac University of Iowa pre-post result (burnout 4.16 to 3.16, P=0.005) and the category-wide caveats: high omission and hallucination rates in simulated settings, small samples, publication bias (accessed 2026-08-12)
- nabla.com/security — Official security page: SOC 2 Type II and ISO 27001 certified, Google Cloud hosting with region choice at organisation creation, AES-256 at rest, TLS in transit, annual third-party penetration testing (accessed 2026-08-12)