The AI landscape moves fast. Models that were state-of-the-art six months ago have already been replaced by faster, more capable, and more cost-effective successors. For clinics running patient-facing chatbots and Medical Agents on A2V2.ai, staying current with the best available models should not require constant manual oversight.
That is why we handle it for you.
This month we ran a full catalog refresh. A number of older models have been retired, several newer and more powerful ones have taken their place, and the list of HIPAA-eligible models available for Medical Agents got significantly bigger. If your chatbot was still configured with a retired model, we have already moved it to the recommended replacement. No downtime. No action required on your end.
Here is the full breakdown.
What Changed at a Glance
Three big moves happened in this update.
First, OpenAI's older model families have been retired. The GPT-4.x line (GPT-4.1, GPT-4o, GPT-4.5) and the o-series models (o3, o3-mini, o4-mini) are gone. The GPT-5 family is now the default across the platform.
Second, the HIPAA-eligible model lineup expanded dramatically. Previously, Medical Agents were limited to a handful of open-source models like Llama, Qwen, and GPT-OSS. Now, the Claude 4.6 family from Anthropic (Opus, Sonnet, and Haiku) and Google's Gemini 2.5 (Pro and Flash) are all HIPAA-eligible. So is GLM-5. That means clinics running Medical Agents now have access to flagship-tier reasoning models inside a fully compliant environment for the first time.
Third, most credit costs went down. Claude Opus 4.6 dropped from 10 credits per message to 3. Claude Sonnet 4.6 dropped from 3 to 2. GLM-5 dropped from 2 to 1. The one increase is Gemini 2.5 Pro, which went from 1 credit to 2 credits per message.
Models Retired and Their Replacements
If you never explicitly picked a model, the platform default moved from GPT-4o-mini to GPT-5-mini. For Medical Agents, the default is now Llama 3.1 8B Instant.
New Models Now Available
From OpenAI: GPT-5, GPT-5-mini, GPT-5-nano, plus GPT-5.4, GPT-5.4-mini, and GPT-5.4-nano.
From Anthropic: Claude Opus 4.6, Claude Sonnet 4.6, and Claude Haiku 4.5. All three HIPAA-eligible.
From Google: Gemini 2.5 Pro and Gemini 2.5 Flash. Both HIPAA-eligible.
Also new: GLM-5. HIPAA-eligible, 1 credit per message.
HIPAA Coverage Just Got Wider
This is the biggest change for clinics running Medical Agents.
Until this update, the HIPAA-eligible model selection was limited to open-source models. Llama, Qwen, and GPT-OSS are solid models, but they do not match the reasoning depth of flagship commercial models. Clinics that needed the best available AI for complex patient interactions had to choose between capability and compliance.
That trade-off is gone.
With Claude Opus 4.6, Claude Sonnet 4.6, Claude Haiku 4.5, Gemini 2.5 Pro, Gemini 2.5 Flash, and GLM-5 all now HIPAA-eligible, Medical Agents can run on the same caliber models that power the most advanced AI applications anywhere, all within a fully compliant environment.
For clinics managing complex longevity protocols (NAD+ therapy, peptide therapy sequences, hormone optimization programs, rapamycin cycling, senolytics therapy), this means Medical Agents that are designed to handle nuanced patient questions with significantly more depth and accuracy.
Your AI Is a HIPAA Violation Waiting to Happen explains the compliance landscape in detail if you want the full picture.
Auto-Migration: What We Changed and What We Did Not
We touched one thing: the model selection field. That is it.
What we did NOT touch: knowledge base, chat history, instructions, temperature settings, sub-agents, skills, tools, CRM data, parameters, and prescriptions.
Important for Medical Agent users: HIPAA compliance was preserved throughout. If a retired HIPAA model's replacement was non-HIPAA, we substituted the HIPAA default (Llama 3.1 8B Instant) instead.
No in-app notification for the migration. Check your Sandbox, then Choose Your GPT picker to confirm. Contact support@a2v2.ai if anything looks wrong.
Credit Changes Worth Knowing About
If you are running high-volume on Gemini 2.5 Pro, consider Gemini 2.5 Flash or Claude Sonnet 4.6 as cost-effective alternatives.
Will My Chatbot Behave Differently?
Probably yes, but subtly. Newer models tend to be more concise, stronger on complex reasoning, and slightly different in tone. For most chatbots, these will be improvements. But if you fine-tuned instructions tightly around the old model, spend a few minutes in the Sandbox running common prompts and tweaking if needed.
For Medical Agents previously on open-source HIPAA models, this is a real opportunity. Claude Opus 4.6 and Gemini 2.5 Pro are noticeably stronger on long-document reasoning and complex protocol questions.
What You Should Do
Never picked a model: Nothing. New defaults are running.
Explicitly picked a retired model: Open Sandbox, confirm replacement, run test prompts, adjust instructions if tone shifted.
Running Medical Agent on older HIPAA model: Try Claude Sonnet 4.6 or Gemini 2.5 Pro in Sandbox with real patient questions.
Price-sensitive on Gemini 2.5 Pro: Consider Gemini 2.5 Flash or Claude Sonnet 4.6 at lower or equal cost.
Questions about migration: Email support@a2v2.ai.
Why This Matters for Longevity Clinics
The language model powering your chatbot is the brain behind every patient interaction. Better models mean better patient interactions. And better interactions are correlated with stronger adherence, higher retention, and improved outcomes.
This update ensures every A2V2.ai user has access to the most capable, cost-effective, and compliant AI models available. No manual upgrades. No compliance gaps. No downtime.
Read more: The Silent Revenue Killer, Why Your NAD+ Patients Quit, and AI for Longevity Clinics.
