smolsiby Happiest Labs
FAQRequest evaluationEvaluate

Smol super intelligence

Frontier intelligence.
Smol enough to own.

A complete inference system, running on hardware you own. Local execution. Direct answers. No token metering.

Request an evaluation Explore the system
Private beta Now evaluating

Engine. Index. Tools. Workers.

From Happiest Labs

01 / The inference stack

Smol by design.
Built layer by layer.

Hardware you own, sized to the work. The foundation of a system that can run with the network unplugged.

An in-house inference engine tuned to its hardware. Certified open-weight models, direct answers, and vision in the same session.

Documents, spreadsheets, and images become a searchable local index. Only the relevant material enters the model’s context.

The model specifies the analysis. The host executes the calculations and renders the tables and charts.

Background workers handle indexing, curation, and delegated analysis while the foreground session stays responsive.

Engine, index, tools, and workers. One local system, with your hardware and your information at its foundation.

Your silicon 01 smolsi engine 02 Document index 03 Tool surface 04 Workers 05 smolsi, assembled 06
smolsiEngine · Index · Tools · Workers
Local by constructionSystem architecture
01 / 06Scroll to assemble

02 / Capabilities

Local intelligence. From question to computation.

FIG. 01 YOUR QUESTIONLOCAL MODELANSWER

01 Chat & vision

Direct answers and vision on the local engine, with your information in context.

Ask directly
FIG. 02DOCUMENTSSPREADSHEETSIMAGESLOCAL INDEX

02 Document index

Search documents, spreadsheets, and images in a local index. Only relevant material enters the context.

Bring your context
FIG. 03MODELHOSTSPECIFICATION

03 Workbook & tools

The model specifies the analysis. The host executes the calculations and renders the tables and charts.

Compute the answer

Local model families

  • Qwen
  • Llama
  • DeepSeek
  • Mistral
  • Gemma
  • Kimi
  • gpt-oss

Qwen is certified for the current release. Additional families require certification before deployment. Discuss your model ↗

03 / Agent workflows

Agents that work inside your perimeter.

Find context, delegate analysis, and execute tools with a local model. The work stays on your hardware, even with the network disconnected.

Air-gapped deployment Model, data, and tools stay local
  1. 01 / Context

    Find the relevant material.

    Search local documents, spreadsheets, and images. Bring the useful context into the task.

  2. 02 / Reasoning

    Choose the next step.

    The model directs the task and delegates analysis to background workers.

  3. 03 / Execution

    Run the work locally.

    The host executes calculations and builds the tables and charts in your workbook.

  4. 04 / Review

    Bring the result back.

    Inspect the analysis and workbook, check the underlying material, and decide what happens next.

Example workflowCompare supplier proposals and build a cost breakdown.

Evaluate your workflow

Reading this with an agent?

Open the machine view Raw Markdown ↗

04 / Deployment

Your hardware. Your perimeter.

One team evaluates the workflow, configures the system, and supports the deployment.

Concept: one workstation with its own local compute.

Workstation

One analyst. A signed desktop application.
Works with the network unplugged.

Available
Concept: three workstations connect to one shared server.

Shared node

Local inference for a team, with concurrent sessions and background workers.

Available
Concept: connected servers inside a private perimeter.

Private cluster

Dedicated capacity for isolated environments, with deployment engineering.

Early access

Per seat or per node.
Never per token.

Annual licenses. Pricing on request.
Hardware and services sized to your deployment.Request an evaluation

Private by construction

For work that needs to stay inside your perimeter.

Offline execution, local document processing, and deployment controls configured around your organization’s requirements.

Security & assurance

Deployment options cover data residency, access control, audit logging, encryption, and network isolation. Every ingested file is scanned for prompt injection, with the verdict stored alongside the document. Production model directories carry byte-level manifests and release stamps.

Supporting documentation for security reviews is available under NDA. Request security documentation ↗

Compliance & assurance.

For regulated work, inside your perimeter.

  • ITAR
  • CMMC
  • SOC 2
  • HIPAA
  • GDPR
  • GLBA
  • NIS2

Supporting documentation and applicable scope are available under NDA during your security review.

Request documentation

05 / Get started

Bring one real workflow.

Your documents. A configuration suited to the task. An evaluation with the engineers who build the system.

An engineer replies within one business day.

Prefer email? hello@smolsi.com

A few useful details

Which models does smolsi run?

Local model families include Qwen, Llama, DeepSeek, Mistral, Gemma, Kimi, and gpt-oss. Qwen is certified for the current release; additional families require certification before deployment. Ask about the model and configuration you need during your evaluation.

What hardware do I need?

We size the configuration around your workflow. Workstation and shared-node configurations are shipping; private clusters are in early access. Hardware selection and procurement are available through our deployment team.

How is smolsi licensed?

Annual licenses per seat for workstations or per node for shared inference. Enterprise deployments are quoted. There is no per-token metering; pricing is shared during your evaluation.

Can it run without a network connection?

Yes. The model, document index, and tools can operate with the network disabled. We configure deployment controls around your requirements and provide supporting security documentation under NDA.