NeuroCortex.
India's 220B multimodal AI engine.
Built in Bharat by Dr. Nitnem Singh Sodhi at Bharat NeuroTech. Approximately 220 billion parameter sparse Mixture-of-Experts (~15B active per token). Natively across text, vision, audio and documents. Same engine powers the Lab chat, the ₹1,799 AI Audit, the consult prep notes, and every Custom AI Deployment we ship.
Two engines, one INR wallet. Switch between NeuroCortex and EQCortex (~580B, text-only reasoning) from the model dropdown in any chat. Plug either into an MCP-compatible client via /mcp.
What's under the hood.
| Engine | NeuroCortex · by Bharat NeuroTech |
| Parameters | ~220 billion (sparse Mixture-of-Experts, ~15B active per token) |
| Context window | Infinite Context Window · shared with EQCortex · up to 100B tokens/request (enterprise) |
| Modalities | Text · Vision · Audio · Documents (PDF/DOCX) · Short video |
| Reasoning | Standard · NeuroReason (Deep Thinking) · NeuroAgent (tool-use) |
| Inference | INR-billed, paise-metered |
| Data handling | DPDP-aligned, audit-trail, session-only specs |
| API surface | OpenAI-compatible REST · Bearer auth · per-key daily caps |
| SLAs | Upstream timeouts: 60s non-stream · 120s stream · auto-refund on failure |
Backend infrastructure is proprietary. What we publish is capability, architecture class, price and audit trail — the engine identity, prompting, tool-runtime, audit trail and billing are all ours.
NeuroCortex vs EQCortex — pick the right engine.
Real-time multimodal chat, decoding, vision, voice and document understanding. ~15B active parameters per token, so latency stays chat-grade. Default for Human Intelligence, Business Intelligence and everyday audit chat.
~580B proprietary architecture, ~60B active per token. 800K-token native window, scaled to virtually infinite via NeuroMemory. Purpose-built for emotional intelligence, long-horizon planning, agentic tool-use, code synthesis. India's answer to Mythos and other frontier reasoning models. Use for hard, multi-hour problems where quality > latency. Read the EQCortex spec →
One engine, four modalities.
Long-form drafting, summarisation, code, policy analysis. Toggle Deep Thinking Mode to reason privately and return only a clean, defensible answer.
Upload photos, screenshots, diagrams or scanned documents. NeuroCortex reads them inline and reasons over them in the same turn.
Voice notes, calls, short clips — transcribed and analysed per minute via /api/neurocortex/listen.
PDFs and DOCX up to the Infinite Context window. Powers the ₹1,799 AI Audit chatbot and the consult prep workflow.
India's only AI engines with a virtually infinite context window.
Powered by NeuroMemory — the Memory layer of Dr. Sodhi's Software Neurotechnology paper (IndJCST 2026, DOI 10.59256/indjcst.20260502087). Open access. Read the paper →
NeuroCortex has a 200K native context window. Software Neurotechnology (NeuroMemory) scales it to virtually infinite. Tier ceilings per request:
| Surface | Ceiling per request |
|---|---|
| Free trial (no signup) | 1M tokens |
| Signed-in chat | 10B tokens |
| Public API | 25B tokens |
| Enterprise / custom deployment | Up to virtually infinite |
Inputs up to ~32K tokens are included in the base ₹0.50/msg. Larger inputs pay a size-banded Extended context surcharge — see Pricing. NeuroMemory is proprietary; implementation is closed, capability is not.
Paise, not pennies.
Same INR wallet for both engines. EQCortex bills at the same base rate. API calls debit the same wallet at a tier multiplier — see /api/docs.
12 standards. Flat fee per guided run. ISO 42001/27001/27701 lead-auditor upgrade.
One hour, NeuroCortex on screen, written decision map + 24-hr async follow-up.
The peer-reviewed 20-layer framework NeuroCortex ships (Sodhi, IndJCST 2026).