? India just told the world’s biggest AI labs to hold its chai.
A Mumbai-headquartered AI firm called Fractal Analytics walked into New Delhi’s prestigious India AI Impact Summit 2026 on February 19, 2026 — and walked out with a claim that has the global AI community doing a double-take. Their healthcare AI model, Vaidya 2.0, scored 50.1 on OpenAI’s HealthBench (hard) — making it the first AI model in the world to cross the 50+ threshold on that benchmark, surpassing both OpenAI’s GPT-5 and Google’s Gemini Pro 3. ???
Yes, you read that right. An Indian company just out-AI’d the AI giants — on their own benchmark.
What Even Is HealthBench (Hard)?
Before you dismiss this as corporate PR noise, let’s understand why this benchmark matters.
OpenAI’s HealthBench is not your standard multiple-choice medical exam. It evaluates AI models on realistic, complex healthcare conversations — the kind doctors actually have: ambiguous symptoms, high-stakes decisions, multi-step reasoning under uncertainty. The “hard” subset specifically tests the most demanding clinical reasoning and health data tasks — the scenarios where getting it wrong has real consequences.
Most top models were hovering below 50 on this benchmark. Vaidya 2.0 crossed it at 50.1. That’s not a marketing rounding — that’s a genuine first-in-world milestone, according to Fractal.
Suraj Amonkar, Fractal’s Chief AI Research & Platforms Officer, put it plainly: “By ranking top globally on the OpenAI HealthBench (hard) that focuses on realistic healthcare conversations… Vaidya 2.0 models demonstrate a health-operating system approach that is not only more comprehensive, but also more reliable and accurate across a wide variety of healthcare workflows.”
Not Just a Chatbot — A Healthcare Operating System
Here’s what separates Vaidya 2.0 from generic AI assistants that can recite drug names from Wikipedia. Fractal calls it a “Healthcare Operating System” — and that framing is intentional.
The model is built using post-trained reasoning and agentic architectures, meaning it doesn’t just answer questions — it can take multi-step actions integrated into real healthcare workflows. Think of it less like asking Google a health question and more like having a specialist who actually acts on what you tell them.
Recommended Product
Amazon Renewed — Best Deals of the Day
🛒 View on Amazon →As an Amazon Associate, we earn from qualifying purchases. Price and availability may vary.
The real-world use cases already built into Vaidya 2.0 include:
- ? Emergency Assist — Rapid triage and decision support during critical windows, when seconds count in an understaffed ER
- ? Symptom Checker — High-fidelity wellness reasoning for citizens, designed to handle India’s diverse health scenarios
- ?? Patient Journey Assist — End-to-end support from first symptom to treatment adherence — bridging the notorious follow-up gap in public healthcare
- ???? Doctor Assist — Support for medical professionals backed by MedExpert benchmark performance
- ?? Administrator Assist — Health data task support for hospital management workflows
That’s not a product — that’s an ecosystem.
The India Angle: Ancient Vaidyas, Modern Code
This is where it gets genuinely exciting — and a little goosebump-inducing — for anyone rooting for India on the global stage. ?
Fractal’s Co-founder, Srikanth Velamakanni, captured the significance at the India AI Impact Summit: “India has built strong digital health foundations over the past decade — from ABHA health IDs to Ayushman Bharat and e-Sanjeevani. The next step is adding reliable reasoning intelligence to those systems. When you combine India’s digital health infrastructure with reliable reasoning AI, you unlock a new operating model for public health.”
This isn’t just product launch talk. Fractal is a selected partner under India’s ?10,300+ crore IndiaAI Mission — a government programme specifically designed to build India’s sovereign AI capabilities so the country isn’t perpetually dependent on American or Chinese models.
Vaidya 2.0 is positioned as the first of many vertical foundation models for the Global South — frugal, scalable, built for impact in resource-constrained settings rather than luxury Silicon Valley offices. The integration potential with Ayushman Bharat Digital Mission and e-Sanjeevani (India’s national telehealth platform serving millions of rural patients) makes this far more than a benchmark achievement — it could, in theory, reshape how 1.4 billion people access healthcare guidance.
But Let’s Be Real — Questions Remain ?
No honest analysis skips the caveats.
The benchmark is healthcare-specific, not general intelligence. Vaidya 2.0 beating GPT-5 on HealthBench hard doesn’t mean it beats GPT-5 on everything — or even on all healthcare tasks. Benchmarks are snapshots, not report cards.
Real-world performance is unproven at scale. How does it handle queries in Bhojpuri, Marathi, or Tamil? How does it perform in low-bandwidth rural settings where e-Sanjeevani operates? How does it handle India-specific diseases, regional medication names, or Ayurvedic context? These questions don’t have public answers yet.
Regulatory clearance is a separate mountain entirely. Any AI tool deployed in clinical settings in India requires CDSCO (Central Drugs Standard Control Organisation) approval. That process is long, rigorous, and as it should be. A benchmark score doesn’t equal a hospital deployment.
AI hallucination in diagnostics is existential risk. Healthcare is the single domain where a confidently wrong AI answer can cost a life. Fractal’s “reliable and accurate” claims need independent validation, not just self-reported benchmarks.
None of this is to diminish what Fractal has achieved — it’s genuinely impressive. It just means the real test happens outside the benchmark lab.
Why This Story Is Bigger Than It Looks Right Now
As of this writing (February 20, 2026), Vaidya 2.0 is barely 24 hours old in public consciousness. The mainstream tech media hasn’t fully caught up. There are no deep Reddit debates yet, no independent researcher teardowns, no viral demos. This is the calm before the storm.
Fractal’s stock reportedly surged ~4% post-announcement — suggesting the investment community has already noticed. But the broader conversation — about what it means for India’s healthcare system, for AI ethics, for doctor-patient relationships, for the IndiaAI Mission — that conversation is still waiting to start.
India produced the world’s first plastic surgery treatise (Sushruta Samhita, ~600 BCE). We systematized surgery, yoga, and holistic medicine millennia before modern hospitals existed. The word “Vaidya” literally means one who has knowledge — derived from Sanskrit Veda, the root of wisdom.
Naming this model Vaidya isn’t corporate branding. It’s a statement of intent. ?
What You Should Watch Next ?
- ? Explore Vaidya 2.0 directly at vaidya.ai
- ? Read OpenAI’s HealthBench methodology to understand what 50.1 actually means in practice
- ? Track how e-Sanjeevani and Ayushman Bharat Digital Mission integrate with third-party AI tools
- ? Catch our full video breakdown on YouTube — we deep-dive the benchmark numbers and what they mean for everyday Indians ?
If India’s AI moment has arrived, Vaidya 2.0 might just be its first cry. ? Subscribe and hit the bell — because this story is only going to get louder.
