Gen-1

What is Gen-1?

Gen-1 turns a single request into a presentation and is Genspark's first in-house model, purpose-built for AI Slides Standard mode. Gen-1 is the model, AI Slides is the product, and Standard and Ultra are the run modes.

Gen-1 is built from MiniMax M3, an open-weight model, and was post-trained with reinforcement learning inside the same production system that runs AI Slides. For the benchmarks and method, see the launch post.

Where you'll find Gen-1

Gen-1 is the Standard model inside AI Slides for building your deck. New decks start in Standard unless you've chosen Ultra, or your team admin set Ultra as the default. Nothing to turn on; if you want Ultra, pick it before you generate. The active mode is always shown on the mode pill at the bottom of the AI Slides input area. A deck's mode is fixed once you send your first prompt.

• Standard mode — runs Gen-1 by default (included in every plan; see the vendor note below if your organization restricts AI model vendors)
• Ultra mode — keeps offering frontier models (top-tier models from other labs)

Gen-1 is additive: it joins the models Genspark already orchestrates. Standard mode uses Gen-1 by default, and Ultra mode keeps the frontier models available whenever you want them.

If your organization restricts AI model vendors, AI Slides uses only models from your approved vendors. Standard runs Gen-1 when Genspark is approved, except when Google or Anthropic is the only other AI model vendor approved; then it uses that vendor's model. Without Genspark, Standard uses an approved Google or Anthropic model. If neither Google nor Anthropic is approved, the Standard and Ultra options don't appear, and AI Slides uses another approved model or shows a policy message, depending on your admin's settings. Members don't need to change anything. Admins manage this in Team Admin → Connectors → "Restrict AI Model Vendors" (available once it's enabled for your organization).

What you get with Gen-1

Gen-1 specializes in turning a request into a complete deck quickly, so Standard is a fast, economical way to get from a prompt to a full first draft. Use Ultra for high-stakes decks and decks that go outside your team. Ultra still runs frontier models from other labs.

• Polished visual design — built for presentation-quality decks
• Clean, reliable layouts — structured slides built end to end, ready for your audience
• End-to-end slide building — planning, design, and iteration handled from a single prompt

Privacy & your data

Gen-1 runs on dedicated capacity operated for Genspark. Your data is not shared with the base model's developer, MiniMax is not a subprocessor, and MiniMax has no access to your data. This is the same boundary that applies to every model used.

You control whether your content is used to improve Genspark. Manage the AI Data Retention setting in your account settings — it governs both model training and the de-identified evaluation described in the FAQ. Team and Enterprise workspaces are opted out by default.

Cost

Gen-1 is the model used for Standard mode in AI Slides, available on every plan. Using it consumes credits at the Standard rate — see the Credits Guide for what a deck uses, and Membership Plans for what each plan includes.

Commercial use of the decks you generate depends on your plan. See Membership Plans for commercial-use rights, and the Terms of Service for attribution requirements.

For teams wanting API access to this model, reach out to us. For Enterprise workspaces with model restrictions, see "Does Standard work with my organization's model restrictions?" in the FAQ below.

FAQ

Is Gen-1 open-weight?

No. It is built on the open-weight model MiniMax M3, but the post-trained Gen-1 model itself is not open-weight.

Does this mean Genspark is done orchestrating other models?

No. Genspark continues to orchestrate leading models for every task, frontier and open alike. Gen-1 is the first model trained where our scale and product experience give us an advantage — slides lead the way.

Was Gen-1 trained on my decks?

No. Training used only internally constructed tasks. A separate evaluation used 200 de-identified slide tasks from consumer users who had AI Data Retention enabled; Team and Enterprise data was not used.

Does Gen-1 cost extra?

No — it is the default in Standard mode of AI Slides, available on every plan, and consumes credits at the Standard rate like any other Standard-mode generation.

Which should I use, Standard or Ultra?

Use Standard for everyday and internal decks, and for first drafts. It's fast and economical, and you can refine the result with follow-up prompts or Advanced Edit (the Edit button; Professional Mode only). Use Ultra when the deck goes outside your team or when it's high-stakes. For brand colors, attach your brand template or use a saved Skill, then check the palette before you share. Pick the mode before you generate: it's fixed once you send your first prompt. See Standard vs Ultra in AI Slides.

Does Standard work with my organization's model restrictions?

Under "Restrict AI Model Vendors" (available once it's enabled for your organization), AI Slides uses only models from your approved vendors. Standard runs Gen-1 when Genspark is approved, except when Google or Anthropic is the only other AI model vendor approved; then it uses that vendor's model. Without Genspark, Standard uses an approved Google or Anthropic model. If neither Google nor Anthropic is approved, the Standard and Ultra options don't appear, and AI Slides uses another approved model or shows a policy message, depending on your admin's settings. To use Gen-1, ask your admin to approve Genspark in the vendor list, or contact us to turn the list on.

Can I use other models for slides?

Yes. Ultra mode in AI Slides continues to offer frontier models from other labs.

What should I do if my deck doesn't look right?

Describe the change and the slide is regenerated, or edit any slide directly with Advanced Edit (the Edit button; Professional Mode only). You can start a new deck in Ultra mode if you want frontier models.