Multi-model AI workspace

Every model.One volume.

418 models held in a single space — one wallet, one memory, one set of files. Move through it. Then watch a prompt find its own way.

BEST TECH · DEMO NIGHT @ FLORIDA ATLANTIC →
scroll to explore

01 · Split the Diff

Every token saved,
split 50/50.

Nyquest compresses context before it's billed. Whatever that saves against the provider's rate is the diff — and we split it down the middle. Half comes off your bill. Half is how Nyquest gets paid. That's the whole business model: no subscription, ever. We're partners in the savings, and we only earn when you do.

Provider rate for the call · 100%
Tokens compression removes · the diff
Your half of the diff · off your bill
Nyquest's half · our fee — nothing else

Free $0
  • BYOK for any provider, billed direct
  • Every currently-free model in the catalog
  • 20 requests / minute · 10 MB uploads
Pro — one $5 fund, permanent 50/50
  • All 418 models unlocked, forever
  • Same split as enterprise, from the first $5
  • Wallet pays provider rate minus your half of the diff
  • No monthly fee, no auto-renewal — stays Pro after the balance runs down
  • 120 requests / minute · 100 MB uploads
Enterprise — custom 50/50
  • Same split, at your volume
  • Custom agreements, limits, and support
  • The diff is measured per call and shown on every invoice
  • Talk to us →

One rule from the lowest paid tier to enterprise custom: the savings are split 50/50, and the split is the only thing Nyquest charges. Full pricing →

02 · The catalog

The catalog isn't a list.
It's a space.

Every name you flew past on the way in is a real model in the catalog — hover one back at the surface and it'll tell you its family. Down here, the field yields so the argument can be read.

A dropdown makes you choose. A volume lets you see the shape of the choice — which is the part that actually matters when there are 418 of them and the cheapest one is sometimes the right answer.

AnthropicClaude Fable 5.1 (batch) · Fable 5.1 · Opus 5
OpenAIGPT-6 Astra · Astra (batch) · Astra Pro
GoogleGemini 3.8 Flash (batch) · 3.8 Flash · 3.7 Flash
xAIGrok 4.6 · 4.5 · Build 0.1
DeepSeekV3 · R1 reasoning line
Meta · Mistral · Qwenopen-weight families
Everything elseCohere, AI21, Perplexity, Nous & more+248

03 · How routing works

Send something in.
Watch where it lands.

Nyquest scores every prompt for complexity, sets a cost ceiling, and picks. Sometimes that's the most expensive model in the field. Frequently it isn't — and that's the routing layer quietly earning its keep on every request you send.

Live routing · pick a prompt

Routing figures shown are representative of production behaviour on the current catalog.

04 · Chat, Splicer & Studios

One space.
Not ten tabs.

Chat, generation and verification share the same wallet, the same memory and the same files. Nothing here is a separate product with a separate bill.

Chat

Every model, routed right

418 models behind one selector. Tiered auto-routing picks by complexity and price, or pin any model yourself. Files become context, memory persists across sessions, live web data grounds the answer. BYOK bills direct.

Full catalog →
Create

Every medium, one prompt

Images, natural voice, original music and video — in chat or in dedicated studios, from the same wallet. Describe what you want; Nyquest detects the creative intent and routes to the generator that fits.

See real outputs →
Verify

Splicer: ask them all

One prompt fanned out to several frontier models in parallel, judged for agreement, synthesised into a consensus answer — divergences flagged, with receipts for every model's response, cost and latency.

Explore Splicer →

05 · Twelve design languages

It doesn’t have a look.
It has twelve.

Most tools give you a light mode and a dark mode. Nyquest ships twelve complete design languages — a chassis is a typeface, a palette, a density and its own vocabulary, not a colour swap. Switch whenever the work changes; every one runs on desktop and phone.

Pick a room

Quiet paper to deep space

Paper reads like a document. Quant runs a live ticker. Matrix and Apsis put you in the dark with rain and starfields. Courtroom, Atelier, Atlas, Studio, Ridgeline, Garden — each is a different place to think.

All twelve →
It speaks the part

Vocabulary changes too

A project is a Folio in Atelier, a Case in Courtroom, a Mandate in Quant. The interface adopts the language of the work rather than making you adopt the language of the software.

How Prism was built →
Newest

Prism — Nyquest in 2036

The twelfth chassis: dispersion, spectral edges and a future-facing type system. Built on the shared foundation, so it shipped as design rather than as a rebuild.

Try Prism →

06 · What’s included

A workspace,
not a chat box.

Artifacts

Upload 14 file types — they become model context automatically.

Memory

Facts extracted, confidence-scored, recalled by vector search.

Agents

Multi-step runs with live streaming and cost guardrails.

Web grounding

Live page content injected into context. No extra keys.

Analytics

Every routing decision, token and dollar — visible.

Media studios

Dedicated image and audio studios beside the chat.

Smart routing

Tiered defaults, model pinning, automatic failover.

Chassis

12 complete design languages, one workspace.

Infrastructure

Rust backend, daily backups, AES-256-GCM key encryption.

Explore all products →

07 · Real output

See it.
Hear it.

Every clip below was generated from a single text prompt on Nyquest — no edits, no retouching, no second tool.

Video · Google Veo 3.1

"a sweeping cinematic aerial shot flying over a futuristic coastal city at golden hour, sleek glass towers, calm ocean, smooth camera motion, warm light"

Music · Google Lyria Voice · Gemini TTS

"a bright uplifting electronic track with a warm synth melody" · "Welcome to Nyquest…" (voice: Puck)

Image · Imagen 4 AI-generated isometric creative studio
Image · Imagen 4 Ultra AI-generated photorealistic mountain lake at dawn

08 · The engine

Built in the open.
Build on it.

An OpenAI-compatible API with the same routing, wallet and catalog as the app — and the engine core it grew from is published under MIT at v3.2.8. Full Rust, drop-in proxy, Docker image on GHCR.

Point an OpenAI or Anthropic client at localhost:5400 and ship. No telemetry, no phone-home, prompt content never logged. Run it air-gapped if you want to.

532routing rules
<1msP50 hot path
~3,730×cache hit
MITv3.2.8 release
# one call, any model in the field curl https://api.nyquest.ai/v1/chat/completions \ -H "Authorization: Bearer nq-v1-…" \ -d '{"model":"claude-sonnet-5", "messages":[{"role":"user", "content":"hi"}]}'
# or run the whole engine yourself git clone github.com/Nyquest-ai/nyquest-rust-fullstack-3.2.0 bash ~/nyquest/install.sh

09 · Get started

Come inside.

Free with your own keys, or any model currently available at no cost. Fund the wallet for everything else.

Open Nyquest