AI sovereignty

An open-model NotebookLM alternative

NotebookLM runs on Google's models only. Scholarly runs the same source-grounded workflow on open-weight models from Mistral AI, Z.ai, and Moonshot AI, alongside five other labs, so your notebook is not tied to one vendor's roadmap.

Open-weight Mistral, Z.ai, and Moonshot AIEight AI labs in one workspaceZero data retention on open modelsUS or EU hosting on request
Try it on your own documents
Used by 170,000+ learners and professionals worldwide

Quick answer

Two different things get called an open-source NotebookLM alternative

One is self-hosted software you run yourself. The other is a hosted notebook that does not lock you to a single closed model. Scholarly is the second. The application is not open-source software and you cannot self-host it, and we would rather say that here than let you find out after signing up.

  • Open-weight models, in the product today. Mistral Large 3, Mistral Medium 3.5, and Mistral Small 4 from Mistral AI, GLM 5.3 Flash from Z.ai, and Kimi K3 from Moonshot AI.
  • Eight labs, not one. OpenAI, Anthropic, Google, xAI, Mistral AI, Z.ai, Moonshot AI, and Meta are all selectable, so a deprecation or outage at any one of them is a switch rather than a migration.
  • Zero data retention on the open models. The Mistral, Z.ai, and Moonshot AI routes in our catalog are configured for zero data retention and US-hosted inference.
  • Grounded in your sources, whichever model runs. Every answer and artifact is generated from the PDFs, decks, recordings, and notes you uploaded, with citations back to them.
  • Admins set the policy, not each person. On Scholarly Teams, model access controls apply to the whole workspace at once, so an open-model-only policy is one setting.

If you need software you can run on your own servers, Scholarly is not it. If what you actually want is a notebook that never depends on one company's model, that is exactly what this is.

Updated 4 September 2026. Model catalog current as of this date; labs ship and retire models continuously.

Add the sources you already have

Upload PDFs, Office documents, slides, recordings, videos, and notes. They stay the ground truth for everything generated afterwards.

Choose a model, or an open one

Pick per task from eight labs. Choosing Mistral, Z.ai, or Moonshot AI keeps the work on open-weight models without changing anything else about the workflow.

Generate grounded artifacts

Cited answers, research reports, narrated video lectures, podcasts, slide decks, documents, quizzes, and graded exams, all traceable to the source.

What you get that a single-vendor notebook cannot give you

No single point of failure

When one lab has an outage, reprices, or retires a model, you switch model and keep working instead of waiting.

Open weights as a real option

Mistral, Z.ai, and Moonshot AI models sit next to the frontier ones in the same picker, on the same sources.

A hosting conversation you can have

Teams that need every model hosted in the US or the EU can arrange that by email rather than filing a feature request.

One workspace, many artifact types

Cited chat, deep research reports, narrated videos, podcasts, decks, documents, quizzes, and exams from the same library.

Model choice

NotebookLM runs Google's models. Scholarly runs models from OpenAI, Anthropic, Google, xAI, Mistral AI, Z.ai, Moonshot AI, and Meta.

Open weights

Mistral Large 3, Medium 3.5 and Small 4, GLM 5.3 Flash, and Kimi K3 are selectable on any source in the workspace.

Self-hosting

Not available. Scholarly is a hosted product. If you need to run the software yourself, a self-hosted project is the right tool.

What people usually mean by this search

Most of the traffic for this phrase splits two ways. Some people want software they can clone and run on their own hardware, usually for privacy or cost reasons. Others have decided they like the NotebookLM workflow but do not want the whole thing to depend on Google's model roadmap and pricing. The first group wants a repository. The second group wants model independence, and that is a product decision rather than a licence one.

Where Scholarly actually sits

Scholarly is a hosted, closed-source application whose model layer is deliberately open. The catalog carries models from eight labs and includes open-weight families from Mistral AI, Z.ai, and Moonshot AI. You can move a source between models without re-uploading it, and a team admin can restrict the workspace to the models they are comfortable with. What you cannot do is take the application itself and run it on your own infrastructure.

Why open weights matter even when you do not self-host

An open-weight model can be hosted by more than one provider, which means the model you standardized on is not a single vendor's asset to retire. It also means the price of a given capability has a floor set by competition rather than by one company's pricing page. For a team that has built onboarding modules or an enablement library on top of a model, that difference is the difference between a settings change and a rebuild.

The honest limits

Scholarly is not open-source software, is not self-hostable, and does not publish model weights of its own. The open-weight routes in the catalog are configured for zero data retention and US-hosted inference, not EU-hosted inference; teams that need every model hosted in a specific region arrange that directly with us. And NotebookLM is a good product: if one person needs to read a document set and nothing else, Google's tool is a reasonable choice and this page is not trying to talk you out of it.

Keep exploring

Related pages

The sovereignty pillar this sits under, and the team workspace it powers.

AI sovereignty

The full argument for eight labs, open models, and hosting you choose.

NotebookLM alternative for teams

Shared library, roles, reporting, and one bill.

Scholarly vs NotebookLM

The individual comparison, feature by feature.

Scholarly for Teams

Workspace-wide model access controls set once by an admin.

AI chat features

Source-grounded chat with citations, across every model.

Scholarly for businesses

Research, content production, and training on one bill.

Frequently Asked Questions

Is Scholarly open source?

No. Scholarly is a hosted, closed-source application. What is open is the model layer: open-weight models from Mistral AI, Z.ai, and Moonshot AI are available in the same picker as the frontier models, on the same sources. If your requirement is source code you can audit and self-host, Scholarly does not meet it.

Can I self-host Scholarly?

No. There is no self-hosted build. For teams whose concern is where inference runs rather than where the application runs, we can set up US-only or EU-only model hosting for a Teams workspace; email hello@scholarly.so and we will arrange it.

Which open-weight models can I use?

Mistral Large 3, Mistral Medium 3.5, and Mistral Small 4 from Mistral AI, GLM 5.3 Flash from Z.ai, and Kimi K3 from Moonshot AI. They sit alongside models from OpenAI, Anthropic, Google, xAI, and Meta, and any of them can be pointed at the same uploaded sources.

Does NotebookLM support other models?

NotebookLM is built on Google's own models and does not offer a model picker across other labs. That is a deliberate product choice on their side rather than a shortcoming, and it is the single clearest difference between the two products.

Where does inference run for the open models?

The Mistral, Z.ai, and Moonshot AI routes in our catalog are configured for zero data retention and US-hosted inference. Region-specific hosting for an entire team, including EU-only, is arranged case by case by contacting us.

Does changing model change the output quality?

Yes, models differ, and a smaller open model will not match a frontier model on hard reasoning. What does not change is grounding: every artifact is still generated from your uploaded sources with citations, whichever model produced it.

Can an admin force the team onto open models only?

On Scholarly Teams, model access controls apply to the whole workspace at once, so an admin can restrict which labs and tiers are in play for every seat rather than relying on each member to choose correctly.

A notebook that does not depend on one lab

Bring the sources you already have and pick the model that suits the task.

Trying your first source? Start Free. Working with new material each week? Premium includes 10 AI Creation Credits a week. Need more capacity? Compare Laureate.

Free

Try the workflow with your first source.

$0

/month

Free forever

No card required.

Get Started

  • 3 lifetime AI Chat messages

  • 1 free AI Creation Credit total

  • 2-minute podcasts and video lectures

  • 1 free file upload total (8MB)

  • 5 lifetime quiz questions

  • 1 completed practice exam total

  • 15 lifetime voice minutes

  • 32-page PDF to flashcards

  • 500 lifetime autocomplete words

Use it to generate flashcards, improve a deck, or create a podcast, video lecture, slides, infographic, mind map, study guide, worksheet, spreadsheet, story book, timeline, SOP, flowchart, lesson plan, or outline — or run Deep Research or turn a recording into AI Meeting Notes.

Premium

Most Popular

For ongoing learning and work with new material.

$12

/month

Billed yearly at $144 · save $216

Get Premium

  • Everything in Free, plus:

  • 10 AI Creation Credits per week

  • Unlimited AI chat & autocomplete

  • Podcasts up to 60 minutes, videos up to 45 minutes

  • Unlimited file uploads (up to 300MB)

  • Unlimited study sessions

  • Unlimited exams & quizzes

  • 1000-page PDF to flashcards

  • Combine up to 10 sources per creation

  • Export to Anki

  • Priority support

Laureate

For heavier workloads and larger source collections.

$99

/month

billed monthly · cancel anytime

Get Laureate

  • Everything in Premium, plus:

  • 80 AI Creation Credits per week

  • Podcasts up to 60 minutes, videos up to 45 minutes

  • Unlimited uploads up to 1 GB, plus 3,000-page PDFs

  • Combine up to 20 sources per creation

  • 12 hours of voice mode per day

  • Priority support and early access to new Frontier models

Pricing shown in USD by default. Local currency may be available in app or at checkout.

Compare plans

AI chat

Premium and Laureate include unlimited AI chat. GPT 5.6 Sol and Claude Opus 5 are reserved for credit-metered AI creations.

Free
3 total
Premium
Unlimited
Laureate
Unlimited

AI Creation Credits

AI Creation Credits share one balance across every creation tool. Most models use 1 credit, Sonnet and Kimi use 2, and Frontier models Sol and Opus use 3.

Free
1 total
Premium
10/week
Laureate
80/week

Podcast length

Free episodes are fixed at 2 minutes. Paid plans can set a target length up to 60 minutes.

Free
2 minutes
Premium
Up to 60 minutes
Laureate
Up to 60 minutes

Video lecture length

Free lectures with a length control are fixed at 2 minutes. Paid plans can set a target up to 45 minutes. Short uses its own short video format.

Free
2 minutes
Premium
Up to 45 minutes
Laureate
Up to 45 minutes

File uploads

Free
1 total (8MB)
Premium
Unlimited (300MB)
Laureate
Unlimited (1 GB each)

PDF to flashcards

Free
32 pages
Premium
1000 pages
Laureate
3000 pages

Sources per creation

Free
Up to 1
Premium
Up to 10
Laureate
Up to 20

Practice questions

Free
5 total
Premium
Unlimited
Laureate
Unlimited

Practice exams

Free
1 total
Premium
Unlimited
Laureate
Unlimited

Voice mode

Free
15 min total
Premium
2 hr/day
Laureate
12 hr/day

Autocomplete

Free
500 words total
Premium
Unlimited
Laureate
Unlimited

Export to Anki

Free
—
Premium
Included
Laureate
Included

Support

Free
Standard
Premium
Priority
Laureate
Priority

What customers say

Scholarly has enhanced the quality of notes and questions I administer to my tutees. Insightful, to the point.

TD Georgas

TD Georgas

Tutor

I have very good experience with Scholarly it summarize me scientific text in hovel academic level with citation. I use this for any therapeutic area . It is great and useful tool.

Diana

Diana

Scholarly user

Scholarly has been a valuable tool for my studies. The AI-generated flashcards and intuitive features make organizing and retaining information much easier.

Briana

Briana

Scholarly user

This app is great for studying for big test. Drop your PDF's in the system and it'll do the trick. You can organize it specifically for your needs.

Kelvin

Kelvin

Scholarly user

Your questions, answered

Is Scholarly free to use?

Yes! The free plan includes core study tools with lifetime allowances: 3 AI Chat messages, 500 autocomplete words, 5 quiz questions, 15 voice minutes, 1 practice exam, one free AI Creation Credit total, and 1 file upload. These allowances do not reset. Upgrade to Premium for 10 AI Creation Credits per week and unlimited uploads.

What uses my free AI Creation Credits?

AI Creation Credits are shared across flashcards, podcasts, videos, slides, AI Documents, Deep Research, and the Transcribe tool. The selected model determines the credit cost. Recording live audio and saving its transcript uses your file upload allowance instead; it does not use AI Creation Credits.

Can I cancel anytime?

Absolutely. There are no contracts or commitments. You can cancel your subscription at any time from your account settings, and you'll keep access until the end of your billing period.

What payment methods do you accept?

We accept all major credit and debit cards through Stripe. Pricing is displayed in USD by default, but local currency is available in the app.

Can I use Scholarly with a class or school?

Yes. For a school or organization-wide rollout, contact us at hello@scholarly.so and we'll help you find the right setup.

What happens when I hit a free plan limit?

You'll see a prompt to upgrade. Your existing work is never lost — limits only apply to new actions. Free-plan allowances are lifetime credits and do not reset. Premium includes 10 AI Creation Credits per week, podcasts up to 60 minutes, video lectures up to 45 minutes, unlimited uploads, chat, autocomplete, quizzes, and exams, plus more daily voice time.