Coming soon

Mesh ships in Phase 7. Mesh 3 (free tier) launches on day one — 3 devices, zero cost, forever.

Download →
Modules / 07 · Mesh
Coming soon · Phase 7

Pool RAM.
Run models
no single device can.

Three laptops with 8 GB each become a 24 GB supernode. Mesh 3 is always free — 3 devices, no expiry, zero cost. Need more? Pick your size, pick your duration, pay only for what you need.

StatusComing soon · Phase 7
FreeMesh 3 · 3 devices · ₹0 forever
PaidFrom ₹15 · 5 devices · 1 hr
TechP2P · exo-based · local WiFi
Download → Mesh 3 ships on day one
// Mesh session · live preview 3 nodes · 24 GB pool · model active
adris.tech MESH · ROOM 472819
8 GB Device A 8 GB Device B 8 GB Device C 24 GB unified pool
Room adris.tech-472819
Pool 24 GB · 3 nodes
Model Llama 3.3 70B
Loading layers... 82%
Layer routing
Layer 28–31→ A
Layer 32–35→ B
Layer 36–40→ C
01 What is Mesh?

Small devices,
enormous models.

Most AI models that actually do useful work — Llama 3.3 70B, Qwen 72B, Mistral Large — need 40–80 GB of RAM. No single laptop has that. Mesh solves this.

Multiple machines on the same WiFi pool their RAM together. Each device holds a portion of the model. Inference runs distributed across all of them. From outside, it behaves like one very powerful machine.

  • Model layers split automatically — no manual setup
  • Built on exo (Apache 2.0, open source)
  • Pure P2P — no cloud, no central server, runs on your hardware
  • Works on MacBook, Windows laptop, Linux desktop — mixed hardware
RAM · BEFORE AND AFTER MESH
Without mesh
Laptop A
8 GB available
8 GB
Laptop B
8 GB available
8 GB
Laptop C
8 GB available
8 GB
✗ Llama 3.3 70B needs ~40 GB — won't run
With Mesh 3
24
GB unified RAM pool
Mesh 3 · free
✓ Llama 3.3 70B fits
02 Mesh 3 · free tier

Three devices.
Always free.

Mesh 3 is included on every adris.tech plan — including Explore (₹0 forever). Connect up to 3 machines on the same WiFi and pool their RAM. No credit card. No time limit.

  • Included on every plan including Explore
  • Up to 3 devices on one mesh session
  • Pool up to ~24 GB RAM
  • No expiry — run indefinitely
  • Fully P2P — runs entirely on your own hardware
  • Mixed hardware: Mac, Windows, Linux all work together

Limitation: Mesh 3 sessions are isolated. Two separate meshes cannot connect to each other. That requires a paid plan with a central node.

MESH 3 · FREE · 3 DEVICES MAX
8 GB Device A 8 GB Device B 8 GB Device C MESH 3 ₹0 forever
03 Extend Mesh · paid

Need more
than three?

Scale to 5, 10, 20 devices — as many as your session needs. Hourly sessions start at just 5 devices. Day, week, and monthly passes start at 10. Pay once, use for the full duration.

  • Hourly: 5 devices · ₹15/hr · ₹3.00/device/hr
  • Daily: 10 devices · ₹59/day · ₹0.25/device/hr
  • Weekly: 10 devices · ₹179/week
  • Monthly: 10 devices · ₹399/month · 45× cheaper/hr vs hourly
  • Extra devices: ₹2/hr · ₹6/day · ₹18/week · ₹40/month
  • Paid users become central nodes — control who joins
EXTEND MESH · 10 DEVICES · CENTRAL NODE
8 GB 16 GB 8 GB 8 GB 8 GB 8 GB 8 GB 8 GB Central node · you
1 hour
₹25
10 devices
1 day
₹59
10 devices
1 week
₹179
10 devices
1 month
₹399
10 devices
// 04 — Free vs Paid · full comparison

Two tiers,
one platform.

Feature
Mesh 3 · Free
₹0
Forever · every plan
Extend Mesh · Paid
₹15/hr
5–30+ devices · any duration
Max devices
Up to 3 devices
5–30+ devices — you choose
RAM pool
Up to ~24 GB (3 × 8 GB)
40–240+ GB — run any model
Pricing
Free — no credit card ever
Pay per session · hourly → monthly
Duration
Unlimited — no expiry
1 hr · 1 day · 1 week · 1 month
Inter-mesh
Not available — isolated only
Connect meshes via central nodes
Plan required
Any plan — including Explore (₹0)
Paid adris.tech plan required
05 Inter-mesh · paid only

Two meshes,
one supernode.

Paid users become "central nodes" — they own a mesh session and control who joins. Two central nodes can connect their separate meshes, creating one massive unified pool.

  • Only paid central nodes can initiate connections
  • Central node A sends a request to central node B
  • B's user sees an accept/reject prompt
  • Combined pool = A's capacity + B's capacity
Graceful migration
Node A (10 devices) joins Node B (20-device plan, 10 used). If A disconnects: B has 10 free slots → A's devices migrate automatically. Nobody drops.
Hard disconnect
Node A (10 devices) joins Node B (10-device plan, fully used). If A disconnects: B has no free slots → A's 10 devices leave the mesh.
INTER-MESH · CENTRAL NODE LINK
MESH A Central Node A 8 GB 8 GB 16 GB MESH B Central Node B 8 GB 8 GB 8 GB paid link
06 How Mesh works

Four steps to
a supernode.

Step 01

Open Models

Open adris.tech desktop app. Go to the Models module. Choose the model you want to run — Llama 70B, Qwen 72B, anything that needs more RAM than you have.

Step 02

Run with Mesh

Click "Run with Mesh" instead of Run. adris.tech starts a node on your machine and generates a room code: adris.tech-472819

Step 03

Share the code

Send the 6-digit room code to friends. They open adris.tech on their machine, click Join Mesh, enter the code — they're in.

Step 04

Model distributes

Layers of the model split across all devices automatically. Each machine holds only its share in RAM. Inference runs distributed, locally — no cloud, no API, no cost.

Your devices,
one supernode.

Mesh 3 is free on every plan. No server, no cost, no expiry. Be the first to know when Extend Mesh launches.