Hippocratic AI prices healthcare voice AI by risk tier, from $2.50 an hour for admin calls to $15
Hippocratic AI's new Polaris 6 Ultra is being read as a technology story. The bigger move is the four-tier, lower-priced packaging that pulls health plans and administrative work into its funnel.
Richard Tang · 4 min read
Hippocratic AI~$1.65B — Hippocratic AI, private, latest valuation ~$1.65B released Polaris 6 Ultra on November 2, 2026, and the under-covered news is commercial: the healthcare voice model now ships in four tiers, Ultra, Pro, Flash and Nano, at lower prices, starting as low as $2.50 per hour for Nano 110. That moves the company from one premium product toward a stack hospitals and health plans can buy at different risk levels, and it widens the funnel from patient-facing clinical calls down into administrative work, where the call volume, and the revenue, live 1.
The pricing tier story: Ultra for clinical calls, Nano for the call center
The tiers split by risk. Ultra runs the full safety constellation for patient-facing clinical conversations; Pro handles complex sub-clinical work; Flash covers high-volume subclinical workflows; Nano is for administrative use cases 1. The company's own site prices them from $15 per hour at the top to $2.50 per hour as low as, roughly a six-fold spread between the clinical and administrative ends 10.
The release also notes increased integration with EHR systems, the electronic health records where clinical context lives, which is what lets the agent hear a patient against their chart rather than against silence 1. Hippocratic sells by the hour of conversation, so cheaper tiers do not dilute the model so much as multiply the addressable volume: a health plan's member onboarding and scheduling calls now have a price point.
Hippocratic says a bigger model now runs faster, not slower
A 671-billion-parameter primary model, trained on more than 323 million patient conversations, sits inside a 7.4-trillion-parameter safety constellation of 30 supervisor models and 142 production monitors, and the stack runs 1.3 times faster than its predecessor 1. The agent plans its next sentence while still speaking, reading the patient's mood mid-call, and holds first-sentence latency at 278 milliseconds 1. On 2,778 clinician-reviewed production calls, misheard medications fell from 1.19 to 0.20 per 100 calls, and television speech in the background was misattributed to the patient fell from 64.8% to under 3% 13.
Company-measured call errors fell across the board under the new speech recognizer
- Before
- After
Data
| Before | After | Change | |
|---|---|---|---|
| Misheard meds | 1.19 | 0.20 | -83.2% |
| Misheard names/numbers | 0.57 | 0.00 | — |
| TV misattributed | 64.80 | 3.00 | -95.4% |
The bet: hospitals pay for a clinical stack, not a rented one
The release is aimed at general-purpose voice AI. OpenAI's Realtime API lets developers build speech-to-speech voice agents over WebRTC and WebSocket, the default building block for customer-facing voice agents 8, and its newer GPT-Live-1 model adds full-duplex conversation for developers at $0.05 per minute 5. Voice AI vendor ElevenLabs~$3.3B — ElevenLabs, private, latest valuation ~$3.3B went further last month, launching Eleven Realtime and absorbing recognition, reasoning and speech synthesis into one managed stack, which improved speed and accuracy but removed the text transcript that enterprise compliance and safety tooling runs on, and its mini model lacks HIPAA-eligible configurations 4.
Hippocratic's counter is that healthcare compliance runs on exactly what those stacks drop: an auditable transcript, logged planning decisions a clinician can review sentence by sentence 1.
Nano undercuts the general-purpose benchmark on price; Ultra still lists at five times it
Data
| Value | |
|---|---|
| Polaris 6 Ultra (clinical) | $15/hr |
| GPT-Live-1 (developer API) | $3/hr |
| Polaris 6 Nano (admin) | $2.5/hr |
The healthcare rivals fighting for the same buyer
Healthcare-specific vendors take different routes to the same customer. Nabla, the Paris-based ambient clinical AI maker, just hired an enterprise-scale CEO and serves over 100,000 clinicians across 190 health systems 6. Ambience Healthcare~$1.25B — Ambience Healthcare, private, latest valuation ~$1.25B now ties its platform fees to measured clinical and financial outcomes rather than seats or tokens 7.
And Assort Health, a voice AI platform for specialty practices that raised $120 million in June at a $1.2 billion valuation, is pushing into health systems on proprietary specialty data 9. The competitive pressure runs both ways: tiered, per-hour pricing from Hippocratic squeezes smaller per-call vendors on cost, while outcome-linked contracting from Ambience pressures everyone on proof.
What to hold against the pitch
Every accuracy and safety figure here is company-measured; the independent review of the announcement notes no published clinical evaluation or regulator determination yet backs the claims 2. And the tier shift is a pricing announcement, not yet revenue 1.
The observable test is specific: whether health plans and providers start running administrative call volume on Flash and Nano, which would show up as adoption announcements and case studies over the next few quarters rather than in any benchmark table 110. If that adoption comes, the tiered stack, not the bedside intuition, is the part of this release that will have changed how the money flows.
Get Health by email, free
AI in medicine, pharma and biotech. In our daily email, every weekday.
