DemoModelsCapabilitiesPricingAboutRequest access
Coming soon · Early access for select companies

Intelligence that runs your business.

Anura is a family of AI models hand-built and trained by Genysis IQ for companies. Agents that write and ship software, serve customers in 25 languages and research the live web — running privately on hardware you own.

Request early access → Watch it work
R1 UltraThe engineer A1The front line A1 LiteThe pocket assistant
0models, one family
0languages served natively
0token context on R1 Ultra
0on a single workstation
0private — your data stays yours
See it work

One model family.
Every desk in the company.

Engineering, support, research and the front desk — Anura takes the work, uses real tools, and reports back.

Illustrative demo
The models

Three sizes. One standard.

Each Anura model is trained by Genysis IQ on its own job, so you run exactly the intelligence the work needs.

Flagship35B MoE · ~3B active

R1 Ultra

The engineer

A deep, agentic coding model. It reads your codebase, plans the change, edits files, runs the tests and fixes what fails — then tells you exactly what it did.

  • Builds, edits, runs and tests software end to end
  • Searches and crawls docs for current APIs, with citations
  • Understands screenshots, diagrams and UI mockups
  • 128K-token working memory with automatic compaction
Reasoning depth
Speed
Runs on one workstation GPU or a 32GB Mac
Everyday4B

A1

The front line

A fast, multilingual assistant for the work every team does all day: answering customers, drafting, summarising and researching the live web.

  • Replies in the customer's own language — 25 of them
  • Agentic web search with sources you can check
  • Reads images, receipts, forms and screenshots
  • Runs on ordinary laptops and servers, even CPU-only
Reasoning depth
Speed
Runs on almost any modern machine
Ultra-light2B

A1 Lite

The pocket assistant

The smallest Anura, built for phones, kiosks and devices at the edge. Instant answers, minimal hardware, and it keeps working when the network doesn't.

  • Fastest responses in the family
  • On-device and offline-capable
  • Multilingual, with optional web search
  • Fits in under 2 GB of memory
Reasoning depth
Speed
Runs on phones, tablets and kiosks
R1 UltraA1A1 Lite
Built forSoftware engineering agentsCustomer-facing & office workPhones, kiosks, edge devices
Size35B mixture-of-experts4B2B
Context window128K tokens32K tokens32K tokens
ToolsFiles, shell, tests, web, visionWeb search, crawl, visionWeb search, vision
LanguagesAll major programming languages25 human languages25 human languages
HardwareOne workstation GPU or 32GB MacAny modern laptop or serverPhones and small devices
DeploymentYour servers or Genysis IQ hostedYour servers, hosted, or embeddedOn-device or hosted
Capabilities

Not a chatbot.
A colleague with tools.

Agentic engineering

Reads the code, makes the change, runs the tests and fixes failures until they pass. Every edit is shown as a diff you can review.

Live web research

Searches, reads and crawls the web for anything current — prices, regulations, documentation — and cites every source it used.

Vision

Understands screenshots, invoices, forms, charts and UI mockups — and can turn a design into working code.

25 languages

Customers write in their language; Anura answers in it. Built for companies that operate across borders.

Private by design

Runs on hardware you own. Your code, customers and documents never leave your network unless you choose.

Long-running work

When a task outgrows its memory, Anura compacts the history into precise notes and keeps going — mid-task, without losing the goal.

Use cases

Where each model earns its keep.

Feature delivery

Turn a ticket into a tested pull request: plan, edit, run the suite, fix, summarise.

Bug triage

Reproduce the failure, find the root cause, ship the fix with a regression test.

Code modernisation

Upgrade frameworks and APIs using today's documentation, read live from the web.

Internal tools

Build dashboards and scripts from a screenshot, a mockup or a sentence.

Customer support

Answer customers around the clock in their own language, with your policies.

Market research

Compare competitors, prices and regulations with sources you can verify.

Office assistant

Drafts emails, summarises documents, reads receipts and forms.

Website assistant

A chat assistant on your site that knows your products and today's news.

Kiosks & front desks

Instant answers for visitors, on a device the size of a tablet.

Field teams

An assistant that keeps working offline in warehouses, sites and vehicles.

In-app assistants

Embed Anura directly in your mobile app — no per-message server costs.

Low-bandwidth markets

Serve customers where connections are slow, expensive or unreliable.

Pricing

Simple, early-access pricing.

Per seat, per token, or a private installation on your own hardware. Regional pricing is available for companies abroad.

Pro
$29/ user / month

For individual engineers.

  • About 300 agent tasks a month
  • CLI, web app and API
  • Web search & vision
Most popularTeam
$49/ user / month

For engineering teams.

  • Higher limits
  • Shared admin & sessions
  • Priority support
API
$0.60/ 1M input tokens
$2.40/ 1M output tokens

Build Anura into your own products.

  • OpenAI-compatible
  • Tools, streaming, images
Private install
$25K– $75K / year

On your hardware, inside your network.

  • + $5K–$15K setup
  • Your data never leaves
  • Dedicated support
Business
$12/ user / month

An assistant for every employee.

  • 25 languages
  • Web research with sources
  • Images & documents
Most popularWebsite assistant
$99– $499 / month

Customer chat on your site.

  • $99 · $249 · $499 tiers
  • Priced by conversations
  • Your branding & policies
API
$0.15/ 1M input tokens
$0.60/ 1M output tokens

Fast multilingual intelligence for your apps.

  • OpenAI-compatible
  • Tools & images
Developer
Freewith daily limits

Prototype with A1 Lite at no cost.

  • API access
  • Community support
API
$0.05/ 1M input tokens
$0.20/ 1M output tokens

The lowest cost per answer in the family.

  • OpenAI-compatible
  • Instant responses
For devicesOn-device license
$0.50– $2 / device / year

Ship Anura inside your hardware or app.

  • Or $5K–$25K / year per product line
  • Offline-capable
  • No per-message cost

Early-access pricing in USD, before tax. Final terms are agreed with each partner company.

Built by Genysis IQ

Hand-built.
Trained for companies.

Anura is designed, built and trained by Josh Clark at Genysis IQ. Every dataset, training run and evaluation is done in-house — verified agent trajectories, identity and safety tuning, multilingual work — so the models behave the way a business needs them to.

We fine-tune proven open-weight foundations into specialists, measure them on real tasks, and deploy them where your data already lives. No black box, no data harvesting: a model you can run, own and trust.

Josh ClarkFounder & engineer, Genysis IQ
01

Trained in-house

Custom datasets and training runs, built and checked by hand.

02

Measured, not promised

Every release is evaluated on held-out tasks before it ships.

03

Yours to run

Deploy on your own hardware, in your own country, under your control.

Limited early access

Anura is coming soon
to a select group of companies.

Tell us about your company and what you'd hand to Anura first. We're onboarding a small number of partners for the first release.

We'll only use this to contact you about Anura.

Request sent.

Thanks — the Genysis IQ team will be in touch.