YNTRHLXION v3.1
Deep reasoning — science & codeNamed for the double helix. Built for problems with a correct answer: proofs, research literature, production codebases. Shows its work, checks it twice.
NEWiViRTH presents the DoubleSlit Project
Every answer starts as a wave of possibilities.
YNTR is how it collapses into one. Six AI models — and the public hardware they'll run inside. Ask something, and instead of one scripted reply, YNTR resolves a spread of possible answers into the one that fits you.
Move your cursor · click to collapse the wave
EnglishHindiGujaratiTamilBengaliTeluguMarathiKannadaMalayalamPunjabiUrduOdiaArabicMandarinJapaneseSpanishFrenchPortugueseSwahili+40 moreEnglishHindiGujaratiTamilBengaliTeluguMarathiKannadaMalayalamPunjabiUrduOdiaArabicMandarinJapaneseSpanishFrenchPortugueseSwahili+40 more
0Models on one spectrum
0Languages supported
0Hardware products
0Time zones on the research team
The DoubleSlit approach
The measurement problem, solved for language.
Every token starts as a wide distribution over what could come next. Your prompt, history and intent are the measurement that decides which response reaches your screen.
stage 01 · superpose
01
Superpose
Pretraining across broad, licensed and public data so the model holds many possible continuations at once, not a narrow script.
02
Interfere
Feedback-driven fine-tuning reinforces the response patterns people want and cancels the ones they don't — like waves adding up or cancelling out.
03
Collapse
At inference, your prompt is the measurement: the model resolves its possibilities into the single response you see, inside guardrails set long before you typed.
A model that only ever gives one answer was never really thinking it through.
The family
Six models. One spectrum.
From precise, particle-like reasoning to broad, wave-like exploration. Pick the one that matches the shape of your problem.
VISION v2.2
Multimodal perceptionSees first — screenshots, charts, handwriting, live video — reasoned over as naturally as text.
iON v3
Flagship general intelligenceLong-context reasoning, tool use, and conversation that holds up across a full working session.
UNION v1.4
Unified multimodalText, image, audio and code in one context window. Reads a whiteboard photo and a spec in the same breath.
SYMBION v1.2
Agentic & collaborativePlans a task, delegates the pieces, checks its own output — works alongside other models and tools.
MINIONS family
Small, fast, everywhereA distilled family that runs on a phone, a browser tab or a Raspberry Pi.
Model finder
What are you building?
Choose a task and we'll point you to the right model.
Beyond the model
Where the intelligence meets the pavement.
iViRTH also designs the physical objects YNTR will run inside. Bench and device details, specs and enquiries live on our dedicated hardware site — nodes.ivirth.in.
Shipping — v1.0
Smart Solar Bench
Public seating that pays its own power bill — solar canopy, ambient lighting and an integrated digital display, no trenching required.
Explore on nodes.ivirth.inIn development
Robotic Totem
An interactive presence for shared spaces — designed to notice people, hold attention and surface information. Concept and prototyping stage.
Follow on nodes.ivirth.inBuild with YNTR
One API. Six models. Zero guesswork.
Every model speaks the same request shape, so switching from a fast MINIONS call to a deep HLXION pass is a one-line change.
- One endpoint routes to any model — pick by name or let YNTR choose the cheapest that can do the job.
- SDKs for Python and JavaScript/TypeScript, plus plain REST.
- Data residency options for teams that need requests to stay in-region.
- Free monthly sandbox quota before you talk pricing.
quickstart.sh
Research
Understanding the model, not just the output.
Interpretability
Tools that let us look inside a running model — so if HLXION reasons its way to an answer, we see the path, and catch it quietly heading somewhere wrong.
Alignment & safety
A separate training round whose only job is narrowing the collapse toward outcomes people want: decline the harmful request, show work on the hard one, say "I don't know" when honest.
Questions
Frequently asked.
YNTR is a family of six AI models built by iViRTH — from on-device MINIONS to the flagship iON, plus specialists for reasoning (HLXION), vision (VISION), unified multimodal input (UNION) and agentic work (SYMBION). It's part of the DoubleSlit Project, with data residency options that keep it independent of any single foreign cloud.
YNTR is a family of six AI models built by iViRTH — from on-device MINIONS to the flagship iON, plus specialists for reasoning (HLXION), vision (VISION), unified multimodal input (UNION) and agentic work (SYMBION). It's part of the DoubleSlit Project, with data residency options that keep it independent of any single foreign cloud.
Start with iON for general work. HLXION for proofs and production code, VISION for images and video, UNION for mixed text/image/audio/code, SYMBION for multi-step agentic tasks, MINIONS for fast answers on a phone or browser tab. Or try the model finder above.
Start with iON for general work. HLXION for proofs and production code, VISION for images and video, UNION for mixed text/image/audio/code, SYMBION for multi-step agentic tasks, MINIONS for fast answers on a phone or browser tab. Or try the model finder above.
Yes — v1.0 is shipping: solar canopy, ambient lighting, integrated display and public-grade construction. The v2.0 AI layer is in active development. Full details live at nodes.ivirth.in.
Yes — v1.0 is shipping: solar canopy, ambient lighting, integrated display and public-grade construction. The v2.0 AI layer is in active development. Full details live at nodes.ivirth.in.
Not yet. It's in prototyping, ahead of field trials and public launch. Images are early concept studies, not a final design. Follow progress at nodes.ivirth.in.
Not yet. It's in prototyping, ahead of field trials and public launch. Images are early concept studies, not a final design. Follow progress at nodes.ivirth.in.
Yes. High- and low-resource languages are trained together from the start — Hindi, Gujarati, Tamil, Bengali, Telugu, Marathi, Kannada, Malayalam, Punjabi, Urdu, Odia and 40+ more — including real-world code-switching.
Yes. High- and low-resource languages are trained together from the start — Hindi, Gujarati, Tamil, Bengali, Telugu, Marathi, Kannada, Malayalam, Punjabi, Urdu, Odia and 40+ more — including real-world code-switching.
iViRTH's long-running effort to build models that hold real uncertainty instead of faking confidence, and give people a say in how it resolves — plus the hardware that carries it into public spaces.
iViRTH's long-running effort to build models that hold real uncertainty instead of faking confidence, and give people a say in how it resolves — plus the hardware that carries it into public spaces.
Request a sandbox key from the Developers section — it includes a free monthly quota. Official SDKs cover Python and JavaScript/TypeScript, with a plain REST API for everything else.
Request a sandbox key from the Developers section — it includes a free monthly quota. Official SDKs cover Python and JavaScript/TypeScript, with a plain REST API for everything else.