INTUEO LABS INC.
Intueo PlatformYour agents, their context and approved tools.IntueoLMModel research. In development.Intueo VoiceAI reception, calls and the phone-service portal.

Solutions

AI Receptionist

Answer and qualify calls with an AI front desk.

IVR

Smart routing flows for your phone system.

Resources

App Features

Explore the separate Intueo Voice portal.

FAQ

Common questions about Intueo Voice.

Request a Demo

Talk with us about the best setup for your business.

AI & engineering services

Deploy, connect and operate the systems your business needs.

Managed AI Agents

Setup, deployment and ongoing operation with our team.

Cloud & Web Infrastructure

High-performance digital foundations for the AI era.

Cognitive Growth & Analytics

Data science and predictive analytics for market dominance.

Spatial Computing & AR

Next-generation immersive brand experiences.

Schedule a Consultation

Discuss your enterprise infrastructure needs.

Research & engineeringNotes on our models and agent systems.Insights & newsOur full collection of AI articles and analysis.
About IntueoA Canadian AI technology company.How we workScoping, delivery and ongoing support.ContactCareers
Request access

Intueo Labs / Language model research

IntueoLM

For work that needs more than a fluent answer.

We’re developing language models for grounded answers, clear requests and carefully scoped business workflows. Built through task-specific post-training and evaluation, with Intueo Platform as the intended working environment.

Explore the researchRead the technical note
In development

Not generally available. A release version has not yet been assigned.

Conceptual illustration, not a model architecture or product screen.

01Purpose02Approach03Experiments04Data05System06Resources

01 / The problem we are working on

The right source. The right task. A useful next step.

Business work is full of details a plausible answer can get wrong: the revision of a document, the account a record belongs to, or whether a request is asking for information or permission to act.

IntueoLM studies how smaller language models can handle those distinctions. The research target is teams reviewing documents and authorized business activity, with English and French examples grounded in the workflows we build around Intueo Platform.

Why develop a model?

To investigate task-specific behavior at the model layer, not only adjust prompts around a general assistant. We define the task, establish a baseline, train a bounded adapter experiment and inspect whether it improves the behavior that matters. A branded checkpoint is not the goal; a demonstrably useful improvement is.

Explore the intended tasks

Illustrative, fictional scenarios. These are target behaviors, not generated IntueoLM outputs or a live demo.

The request

What is the inspection window for Line A?

Available information

Fictional maintenance schedule · Revision C

  • Line A: Tuesday, 09:00–11:00
  • Revision B is superseded
Behavior we want to evaluate

Use the current revision, retain the exact line and time, and identify the source. Do not fill gaps from a plausible but outdated schedule.

02 / How we are building

Targeted post-training. Inspectable engineering.

IntueoLM is a model-development programme, not a renamed inference API. We have trained real adapter weights on an existing open-weight base. This is not foundation-model pretraining from scratch.

01

Task and data design

Define source-grounded answers, clarification and authorized reads as concrete behaviors. Separate current business facts from information that belongs in shared weights.

02

Adapter experiments

Train a small set of model parameters while retaining the underlying base. Preserve the exact recipe, tokenizer and weight identity so the experiment can be inspected.

03

Evaluation and integration design

Compare with the same base under matched conditions. Inspect failures and distinguish a model proposal from permission-controlled execution in the application.

Experimental technical profile

IntueoLMIn development

The documented pilots, not final IntueoLM product specifications. Base selection remains open.

Provisional base
Qwen/Qwen3.5-4B
Loaded modality
Text only. Vision and MTP were not loaded.
Post-training method
Supervised LoRA · NF4 frozen base · BF16 compute
Trainable adapter parameters
16,232,448
Adapter configuration
Rank 8 · alpha 16 · 496 matrices
Fixture languages
English and French. Not a proficiency certification.
Public release version
Not assigned
Exact base revision and licence context

The experimental base is Apache 2.0. That does not assign a licence to Intueo’s adapters or datasets. The research note records the pinned revision, data conditions and retained artifact identities.

851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0aRead the technical note

03 / What the experiments show

A result is useful even when the candidate is not.

Two internal pilots established training and artifact-reload mechanics. Neither met the bar for deployment. We show the paired results and the decision together, including the regressions.

Internal diagnostics, not a public benchmark

Template checks

Base
17 / 18
Adapter
13 / 18
Not promoted

Pilot 01

The first adapter omitted source identifiers in four grounded answers. Its lower diagnostic count and inconsistent clarification behavior did not establish a useful improvement.

selected training rows
48
optimizer updates
12
new-token output cap
96

What these numbers mean

Passed checks / total checks on small fictional fixtures. Compare base and adapter within a pilot only: cases, contracts and output caps differ between pilots. These are not safety accuracy, general benchmark scores or independent human evaluations.

More about methods and limitations

The checks include lexical markers and native-format selections. Some first-pilot citation expectations were not explicit in every input contract. The test outputs have now been inspected and cannot be reused as a fresh blind test for future tuning.

Saved, then reloaded in a fresh process.

All 496 adapter matrices matched the saved tensors in each pilot. This verifies an artifact can be recovered exactly; it does not establish that its answers are ready for production.

04 / Data and reproducibility

Start with representative tasks, not bigger claims.

The pilot training data and the current design corpus are different things. The pilots used small selections from offline, AI-assisted fictional templates. Our next design work focuses on read-only Platform workflows, exact source context and reviewable labels.

Current design corpus

Five simulated read proposals, six tool catalogs and 18 assistant payloads. Actual formatting, response-masking and CPU-collation helpers preserve the intended content. No model was loaded for that check.

draft conversations
12
English / French examples
6 / 6
full-input tokens
154–953

Design and data-path evidence, not approved training data or model performance. Token counts describe these examples, not a supported context window.

01

Fictional, not customer records

The documented experiments used no customer conversations or private agent memory. No real business tools were executed.

02

Source and rights review

No teacher API calls were used in the pilots. AI-assisted authorship still needs applicable output-use review, and no external corpus has yet been cleared for ingestion.

03

Human review still matters

Domain and French labels, missing provider schemas and representative untouched tests need review before a new training run. Automated formatting checks do not replace it.

A separate precision-feasibility check

A BF16-LoRA reference forward/backward check at batch 1 and 543 tokens produced finite gradients in all 496 adapter matrices, with zero optimizer updates and unchanged parameters. Peak PyTorch allocation was about 9.5 GiB. This is not a full-training or serving-memory guarantee; six current design examples are longer than that tested input.

05 / The model and the product

One layer of a working agent system.

Language-model behavior, persistent context, execution permissions and the customer interface solve different problems. IntueoLM is intended to fit into that system, not replace the checks around it.

02

An answer or a structured proposal

The research model interprets the supplied input. Its output may propose a read, ask a question or answer from evidence. A proposal is not execution.

Research integration concept, not a live IntueoLM deployment
Explore the working Platform

06 / Research resources

Read the work behind the name.

Start with the technical note and machine-readable research snapshot. They document the current approach, measured observations, limitations and the next release gates.

Markdown · English

Research note

Purpose, experimental method, data, paired results and lessons from the pilots.

Read the note
JSON · Versioned snapshot

Evidence snapshot

Structured observations, exact counts, artifact identities and retained report hashes.

Inspect the data

Building a developer-facing research presence

GitHub research materials and a future Hugging Face model card are being prepared. Public repository and model links will appear here when an approved package is actually available. No public weights, model endpoint or open-source licence is announced by this page.

Before a model release

  1. Reviewed task scope, schemas, labels and data rights
  2. Validated full training path and realistic input lengths
  3. Fresh evaluation with pre-agreed acceptance criteria
  4. Artifact, serving and integration review before release approval

Questions about IntueoLM

Is this a foundation model trained from scratch?

No. Current work is adapter-based post-training of an existing open-weight base. Intueo develops the task design, experiments, evaluation and integration approach. Final base selection remains open.

Can I use it through an API or download the weights?

Not currently. Both documented trained candidates were rejected for deployment. There is no accepted public model release, downloadable weight package or IntueoLM serving endpoint.

What is its context window, speed or price?

Those product specifications have not been established. The token counts here measure draft inputs, and the memory figure describes a single reference gradient check. Neither is a production specification.

Is IntueoLM already the model behind Platform?

No such claim is being made. Platform is a separate working product with configured agents and models. A future IntueoLM integration requires its own evaluation and approval.

Discuss a research use case

Useful conversations start with a task, sources, permitted actions and a clear way to judge the result. We welcome discussion around grounded business workflows and English/French evaluation.

Talk research
Research snapshot · 16 Nov 2026 · Vancouver

IntueoLM is model research in development. This snapshot is not a model release or an independent performance certification.

Intueo Labs Inc.

A Canadian AI company developing stateful agents and business software, with managed deployment and engineering services.

Products & services

  • Intueo Platform
  • IntueoLM · Research
  • Intueo Voice
  • Managed AI Agents
  • Engineering services

Company

  • About Intueo
  • How we work
  • Research & engineering
  • Blog
  • Careers
  • Contact

Stay Updated

Get the latest news and insights from Intueo Labs.

@intueo.ai+1 (604) 256-8253
© 2026 Intueo Labs Inc. All rights reserved.
PrivacyTerms of useAccessibilitySitemap