NOVEMBER 2026 PUBLIC REVEAL

THE FIRST PUBLIC DOSSIER / 001

You know
the name.
Now meet
the idea.

FredAI is being designed as more than a single conversational model. Today we're opening the first public look at its proposed model families, platform structure and computing options.

Public design disclosure. This is not a claim that the models are trained, benchmarked or available.

FRED / MODEL SYSTEMREVEAL 01
ARCHITECTURE REVEALED08 MODEL FAMILIES
EIGHT FAMILIESONE ASSISTANT BRANDLOCAL AND CLOUDDEVELOPMENT ROADMAP

DISCLOSURE 01 / THE LINEUP

Eight families.
Different ambitions.

One name cannot cover every workload. The proposed FredAI range has eight permanent core family names, with the separate Bit (Paced) family below Ember. Select a core family for its currently disclosed status.

Family numbers above identify the eight permanent core families; they are not a capability ranking. Ember starts the core lineup and Apex is its highest-tier concept. Roles for the other six core families remain unannounced.

Before Ember: Bit (Paced). Bit is a separate, smaller entry family planned for local Instant-only everyday assistance, balanced with extra emphasis on speed. “Paced” means infrequent major updates after initial release. Bit 1 remains in development; no trained weights, measured speed or release availability are being claimed.

DISCLOSURE 02 / THE PLATFORM

One ecosystem.
Multiple ways to use it.

FredAI is intended to be the name of both the wider AI platform and its consumer-facing assistant experience. The diagram shows the proposed product organisation, not deployed infrastructure.

THE UMBRELLAFredAIAI models · Platform direction · Research
01 / ASSISTANT

FredAI

One consumer assistant brand using appropriate models and modes for different tasks.

PLANNED EXPERIENCE
02 / IN-APP COMPUTE

Fred Cloud

Cloud-powered services within FredAI products where a local model is not sufficient or not selected.

PROPOSED SERVICE
03 / DEVELOPERS

Fred API Cloud

A proposed external route for developers to access FredAI cloud models and services.

PROPOSED SERVICE

DISCLOSURE 03 / ACCESS

From your machine.
To the cloud.

Different devices and workloads need different deployment choices. These are the directions being explored rather than services promised at launch.

01 / LOCAL

Open-source. On-device.

Free local model distribution is a design goal. Supported hardware, licences and actual model releases are not yet announced.

USER HARDWARE PLANNED
02 / CLOUD

Power when you need it.

Cloud access is intended for workloads that need remote computing. Usage-based API pricing is envisioned, but no rates have been committed.

REMOTE COMPUTE PLANNED
03 / MODES

Designed to adapt.

Lightweight instant-style modes and additional specialised experiences are part of the vision. SuperFred and UltraFred are proposed overlays.

ADAPTIVE EXPERIENCE IN DESIGN

DISCLOSURE 04 / WHAT REMAINS PRIVATE

A reveal is not
a release.

We can share direction without pretending development is finished. These details are deliberately not announced in November.

01No invented benchmarks.Model scores, parameters and performance need independent evidence.
02No invented availability.Neither model downloads nor cloud endpoints are implied to be live.
03No undisclosed security internals.Credentials, private infrastructure, defences and exploit paths stay private.
04No date-forced product release.Christmas beta remains conditional on verification and release readiness.

NEXT / THE CHRISTMAS PROGRAMME

This is only
the beginning.

01 NOV
Website and architecture revealedPublic information · November broadcast
08 DEC
Birthday event and Christmas countdown15:30 and 16:00 UK time
25 DEC
Christmas finale10:00 UK time · Beta remains evidence-gated