QUANTA

Computing,
born for the model.

For seventy years, computers were organized around the CPU.
We believe the next one is organized around the model.

$ boot 

MODEL-NATIVE ARCHITECTURE · MNA

Compute returns to the center of memory.

The cost of LLM inference is not arithmetic — it is movement.
So we stop hauling data to compute, and move compute into the data.

70%

of energy in typical LLM decode
goes to moving data, not math

10×

of effective-bandwidth headroom
opened by PIM and pooled memory

5

principles define the
Model-Native Architecture

† Order-of-magnitude estimates from public research; see the paper for the full argument.

THE MEMORY WALL

Forty-five years of a widening scissor.

Processor compute sprints; memory bandwidth strolls — and the gap compounds.
LLM inference runs head-first into this wall.

compute ~52%/yr memory BW ~25%/yr
By 2025, the accumulated gap is — this is the memory wall

Growth rates use classic architecture-literature magnitudes (Hennessy & Patterson; Wulf & McKee, 1995); illustrative.

THE ENERGY LEDGER

One fetch pays for a hundred multiplies.

A 64-bit floating multiply costs about 4 pJ; fetching the same 64 bits from off-chip DRAM — 640 pJ. Compute stopped being the bottleneck long ago. Movement is.

One multiply (64-bit FP)
4 pJ
= 160 multiplies' worth of energy
One fetch (64 b · DRAM)
640 pJ

Each lit dot = one multiply's energy (4 pJ). Data: M. Horowitz, ISSCC 2014 (45 nm magnitudes); cross-node is illustrative.

NOETON · EXPLODED VIEW

Scroll to take the Noeton apart.

↓ keep scrolling to explode the five layers · click a legend entry, or see the stack overview →

LLM SYSTEMS

Inference, rescheduled.

Prefill eats compute, decode eats bandwidth — they never belonged on the same chip.

Intent intent Model-OS resolve · schedule Compute Prefill · compute-bound Bandwidth Decode · bandwidth-bound token stream

One request: compute-bound prefill and bandwidth-bound decode flow to different hardware domains

Enter the inference series: four interactive labs →

AGENTS

The agent is the new process.

When the LLM becomes the kernel, everything in the OS deserves a rewrite.

Enter the Model-OS series: the OS's next phase change →

Agent runtime

Scheduling, memory, context management — a crew of agents created, suspended, and collaborating like processes.

Tool orchestration

Tools as syscalls: permissions, sandboxing, and audit built into the runtime, not patched on later.

Intent-driven interaction

Users express intent against context; the machine resolves it into action — the end of command-style interaction.

TRY IT · L5

Hand the Noeton an intent.

No command line, no buttons — click an intent and watch the Model-OS compile it into a plan. (Simulated in-page.)

model-os — intent shell
Click an intent above to begin ▍

FROM HERE TO NOETON

Three steps to the Noeton.

Not science fiction — every step stands on industry facts that already happened.

NOW

Software first

The Model-OS runs as a runtime on today's hardware: agent scheduling, vector memory, tool gateways. Fellow travelers: MemGPT, AIOS, the agent frameworks.

NEXT

Memory-centric racks

Inference servers built on HBM-PIM × CXL pooling, with split prefill/decode scheduling. The groundwork exists: Samsung HBM-PIM, SK hynix AiM, the CXL 3.x ecosystem.

THEN

The full Noeton

The complete machine: tensor dataflow fabric plus intent interface. When decode cost drops another order of magnitude, the interaction paradigm turns the page on its own.

EVIDENCE · NEW IN PAPER V2

The world is moving toward the Noeton.

Rubin CPX built only for prefill, CloudMatrix 384 wiring memory past the CPU, MCP standardized in a year —
we archive the industry's moves against the five principles, one event at a time.

RESEARCH

We publish our judgment.