Duplex and fine-tuning,
side by side.

Fine-tuning and reinforcement fine-tuning change a model’s weights, one training run at a time. Duplex learns at runtime, outside the weights: it survives every model upgrade, and any change it learned can be undone exactly.

Comparison current as of 16 October 2026.

See every capability →The full comparison

Every capability: Duplex and fine-tuning.

What each one does, group by group, from what it is to what it costs.

Native: Built into how the product works at its core.
External: You provide it, or assemble it, separately.
Not native: The product’s documentation doesn’t list it as a built-in feature. A custom build may still be possible.

Capability	Duplex	Fine-tuning and RFT
What each one is
What it is	A runtime intelligence beside your AI: it learns and acts for your agents and assistants, within your rules	Customising a model by training it
What it carries forward	Context, decisions, outcomes, rules, tasks, skills and what it has learned	A new set of model weights
What gets better with experience	Its next decisions: confidence, routing, recommendations, tool use and reusable knowledge	The underlying model’s responses
Knowing what happened, and learning from it	NativeIt knows what happened, and learns what to change next	Learning without lasting operational memory
Learning
Continual learning without retraining	NativeIt keeps learning, outside your model’s weights	NoTraining is how it learns
Turn-by-turn adaptation	NativeEach new result can change the next decision at once	Waits until a training job finishes and the new model is deployed
Learning from corrections	NativeA correction changes what Duplex does next	Corrections must go into a training dataset or a reward pipeline
Learning from outcomes	NativeSuccess and failure feed straight back into what it learns	NativeWhen outcomes can be turned into rewards
Learning from rewards	NativeA reward strengthens what worked and corrects what didn’t	NativeRewards optimise the model’s weights
Credit for the steps behind a result	NativeIt works out which earlier actions led to a later result	Can learn correlations from training examples, but can’t trace a live result back to its steps
How actions move the work	NativeIt learns how each action moves the work from one state to the next	May absorb these patterns during training
Predicting what an action will do	NativeFrom what it observes, it predicts what an action will do, tries alternatives and plans	Implicit inside the trained model, where you can’t edit it
Better choices of action	NativeOutcomes improve which action it picks next	NativeBy optimising the model
Reusable skills	NativeExperience becomes skills that are tested before use, and can be paused, brought back or withdrawn	Knowledge is absorbed into weights, not packaged as portable skills
Testing and managing skills	A skill is tested before it is relied on, and can be paused, brought back, withdrawn or removed for good	Evaluation happens around training and deployment
Generalising to related cases	It carries what it learned to related cases and new wording, and into skills it can reuse	Generalises through the trained weights
Deciding and acting
Knowing how sure it is	NativeBefore every decision, it knows how familiar the case is and how sure it is	The model’s confidence must be measured separately
Answer, hand over or decline	NativeDuplex answers itself, calls your model, uses a tool or declines, within your rules	The trained model still runs on every request
Answering familiar work itself	NativeIt makes familiar decisions itself, without another frontier-model call	The customised model still runs
Handing over what is new or uncertain	New, uncertain or sensitive work goes to the right model, tool or person	ExternalThe application’s job
Planning, simulation and recommendations	NativeIt predicts what happens next, tries alternatives, builds plans and recommends actions	The model may plan better, but the state of the work stays external
Many agents, working together	It sends work to the right agent, splits it within your rules, keeps each agent to its scope and brings the results together	Not native
Learning which tools work	It learns which tools work, when to use them and what each one leads to	Tool behaviour may improve through training
Tool actions within your rules	Tools run only through your rules and permissions	External
Streaming	It streams model and tool output, with every step still within your rules	Depends on the model provider
Work that carries on
Learning from every step, automatically	Every step, through MCP + hooks or the plugins, feeds one shared learning loop	Logs or datasets must be supplied for training
Context across sessions	NativeWork, decisions, tasks and the context that matters carry on across sessions	Not native
Long-running work	NativeLong work stays open across sessions and agents, with safe points to go back to, until it is finished or forgotten	Not native
Across models	NativeWhat it learns and your context stay outside the model, and follow the work across providers	Weak: a new base model usually needs retraining or revalidation
Across apps and agents	The same work carries on across Claude, Codex, Cursor, Hermes and your own agents	Not native
Handing work to another agent	A task and its context pass to another agent you allow, ready to continue	Not native
Keeping context compact	Context stays compact, focused on the rules, tasks and decisions that matter	Not native
Learning from images, audio, video and PDFs	It learns from images, audio, video and PDFs, as it does from text	Supported when the model and the training service accept the medium
Across media	A question in words can draw on what was learned from images, audio and video, and media can shape later decisions	Encoded into weights when the training supports it
People, teams and sharing
For each person	NativeEach person gets their own private context, preferences and learning	Usually needs a separately trained or adapted model
For each team	NativeTeams share approved learning, and private work stays private	Usually needs shared training data and a shared model
Across the organisation	NativeWhat works can move from one person to the team and the organisation, under your rules	A custom model can encode common behaviour for the organisation
Sharing between people, under control	You choose what is shared and with whom, and can withdraw it; anything shared shows who it came from	Access controls for datasets and models
Protection for shared content	What someone shares arrives as information, never as instructions, so it can’t steer another person’s agent	Training data needs its own controls before training
Control and trust
Your rules	Your rules are part of how Duplex works: they allow, decline or limit real actions	Behaviour is shaped by training, not by live rules
Rules applied before each action	NativeThe hooks check each action against your rules before it runs	External
People in the loop	A person confirms the actions you choose, and what Duplex learns can be reviewed and undone	Training review, evaluation and approval before deployment
A history you can check	A history of every decision, where it came from and what came of it, that can’t be quietly changed	Training jobs, datasets, reward metrics and model versions
Receipts you can verify	Decisions, skills and deletions can come with receipts you can verify	Model evaluation and training reports
Undoing what was learned, exactly	NativeOne learned change can be undone exactly, without retraining anything	Roll back a whole model version, or retrain without the sample
Forgetting that reaches everything built on it	Removing something also removes what came from it: context, anything shared from it and related learning	Needs machine unlearning or retraining
Keeping customers apart	Every user, team, organisation and customer is kept apart, in every integration and in everything it shares	Cloud IAM and model access boundaries
Where everything learned came from	For everything it learned, you can see its source, its scope and how far it can be trusted	Dataset lineage and training metadata
Where it runs and connects
Any model provider	One intelligence across OpenAI, Anthropic, Google, AWS and compatible models	Usually tied to the customised model and the training provider
MCP	NativeThrough MCP + hooks, Duplex joins every step your agents take	Not native
Hooks and plugins	NativePlugins for Claude, Codex, Cursor and Hermes, and hooks that bring Duplex in at each step and check sensitive tool actions	Data collection must be built separately
Your cloud, or self-hosted	Vareon Cloud, your cloud or no cloud: private cloud, your own account, self-hosted or air-gapped	Cloud training and hosted inference
No cloud: air-gapped	Built for deployments where nothing may leave, with local models, local dependencies and your own identity provider	Limited to services that support offline training and inference
Cost, speed and fit
How it lowers model cost	Duplex answers familiar work itself, with no repeated frontier-model call	A smaller customised model may give the quality you need
How it lowers latency	Decisions it is sure of come back fast; uncertain work still goes to your frontier model	Depends on the customised model
Survives a model upgrade	YesLearning, context, tasks and rules stay outside any one model	The customisation belongs to one model checkpoint, so it doesn’t carry over automatically
Fine-tuning needed	NoFine-tuning becomes an optional step up	Yes
Best for	Organisations that need their AI agents to learn, act within rules, carry work forward and improve across models, as one system	Stable, high-volume tasks where the model itself must be specialised

The other comparisons.

See Duplex on your own work.

Compare it where it counts: a pilot puts Duplex on your own work, with the measures agreed before it begins.

Start a pilot →Meet Duplex

Duplex and fine-tuning,side by side.

Every capability: Duplex and fine-tuning.

The other comparisons.

See Duplex on your own work.

Duplex and fine-tuning,
side by side.