Hi everyone, Henry from Genspark here. 

We recently launched Gen-1 Slides, our first in-house model trained for Standard mode in AI Slides (Included in Free Quota). I wanted to test the Gen-1 slides against our Ultra tier mode and see what the differences are. In this comparison, I am tracking how long it took to generate the slide deck, how many credits it took, and the overall quality of the final deck.

I used the same prompt for both modes and let each one build an AI industry slide deck, going over trends, adoption, and how it's changing the way we work. The prompt used: 
“Create a business presentation titled "How the World Actually Uses AI in 2026". Cover the latest adoption and usage trends — who is using AI, what they use it for, and how it's changing work — and end with what this means for businesses. Keep it clear and data-driven.”

Gen1-Slides (Standard mode) produced a 10 page slide deck in ~10 minutes from scratch and consumed 244 credits. The final output was a warm and informative presentation that is easy to read, explains how the AI has been adopted across the world, the top uses of AI, and AI governance. 

There are a few areas where the visual layout of the design seems squished or could have better spacing between text blocks but overall I really liked the presentation it gave. The most interesting thing I noticed when generating the slides was it did research on this topic first, asked me clarifying questions on what I wanted, generated a basic slide deck, then re-designed the deck to be more visually appealing.

https://preview.redd.it/86gx5v10z2th1.png?width=2048&format=png&auto=webp&s=54050ab7be1c92a0790bafc5ef836fcfaa9da528

https://preview.redd.it/83pzq8f2z2th1.png?width=2048&format=png&auto=webp&s=1ec6163ee970aa84dd36d45696380d7fdf8c9b43

https://preview.redd.it/6dw8zbb3z2th1.png?width=2048&format=png&auto=webp&s=287e5ea0fe9b2f7b4c9d8fa87695607f9acee766

https://preview.redd.it/z39vlo04z2th1.png?width=2048&format=png&auto=webp&s=ceec60399874b9aa893b0560fcdb0cf4f6827a7f

https://preview.redd.it/qlgzrqi4z2th1.png?width=2048&format=png&auto=webp&s=87d562e332805ad67336fd4fc46bc83bcb6c30fd

Ultra Mode created an 8 page slide deck in ~8-10 minutes and used 766 credits. This deck had a warm tone and was very informative but has a lot more visual elements going on in each slide. Before AI Slides started doing anything, it asked me a few questions to get a better understanding of what I was looking for, then did some deep research before starting the slides. It provided similar information as Gen1 Slides mode but it focused on the AI assistant use case. It did a good job detailing how people use it, its adoption, and what it provides. Ultra mode jumped right into the important slides right away in the presentation without giving much context but provided a recap of everything at the very end. 

https://preview.redd.it/qyfglv10z2th1.png?width=950&format=png&auto=webp&s=e65bdb929f3061a7eb63f22a003d67fc0401aa43

https://preview.redd.it/bxi1ool5z2th1.png?width=950&format=png&auto=webp&s=a6fb60cfde14ebc9d5b3868fdfdd575a93a160f5

https://preview.redd.it/hgpcp736z2th1.png?width=1894&format=png&auto=webp&s=71eb3441335e550fe89c5bb6db2325109bd3fdec

https://preview.redd.it/1eltzhx6z2th1.png?width=1892&format=png&auto=webp&s=04bd888a8c8dbb486551654163f2efd2d4b1b9fe

https://preview.redd.it/iqx9kxg7z2th1.png?width=1912&format=png&auto=webp&s=5d8ac631393c500b330252574a3c9f71717dda9a

Overall, each slide deck had a similar tone and provided its sources but the content of each slide was quite different. I liked how Ultra mode asked me clarifying questions first, had a focus on a specific AI use case, and the visual layout was more eyecatching. I liked how Gen1 Slides re-designed the deck once its first draft was complete, had a more broad information approach to AI in all industries, and had information about governance. I’m happy with both outcomes and to be able to see the comparison side by side and know exactly the credit usage difference between these modes. Let me know what you would like to see next!

Hey everyone! Henry from Genspark here.

We have a big change rolling out to Genspark: a new weekly credit balance for Plus and Pro members. 

What's new: Free Quota
Plus and Pro members now get a weekly balance of free credits called Free Quota that can be used for agents (Super Agent, AI Slides, AI Sheets, and more) using Standard mode, before using your monthly credits. Free quota refills every 7 days and doesn’t roll over week by week.

Agents using standard mode draw on the Free Quota balance first, so your monthly credits stay untouched. When the Free Quota balance runs out, the tasks can continue using your monthly credits. Free Quota does not apply to Ultra mode, it draws on your monthly credits.

Why we're doing this

We’re adding this because we want to open up more of the platform for more of you so you can try new features without worrying about credit burn. AI Chat and AI Image stay free with no credit cost or consume your Free Quota balance on core models until the limits are reached, then it will start using your monthly credits.

To see your Free Quota there is a blue bar above the chat box that reads “Using Free Quota at no credit cost”. When the balance is used up, the bar goes gray and shows when the next refill lands.

What's not included

• Ultra mode on your monthly credits, not Free Quota.
• Design, Code, GenMail, Meeting, Translate, Call For Me, and Workflows aren't covered.
• Video and audio generation use the Free Quota balance.
• The Free Quota balance doesn't roll over, it refills every 7 days regardless of usage.

If you're already a Plus or Pro member
Nothing changes for you through December 31, 2026, as long as your subscription stays continuously active. You still get no-credit AI Chat and AI Image on any models, flagship included, within the existing limits.

I know this is a big change. Let me know if you have any questions regarding this!

Hey everyone! Henry from Genspark here.

https://preview.redd.it/ajyaqbly3soh1.png?width=1145&format=png&auto=webp&s=279d040da053577a105d9619d0b8fd87592bb22b

We just shipped Gen-1 Slides, our first in-house model trained to frontier quality, and it's now the default in AI Slides Standard mode! Gen-1 Slides is built from MiniMax M3 and post-trained with reinforcement learning inside the same production system that runs AI Slides. Ranking first in 8 of 9 internal and public benchmark comparisons putting this in the same tier as Claude Opus 5.

Gen-1 Slides: 

• Gives you a clean, reliable, polished visual design
• Is trained specifically for making presentations giving you the quality slides of a frontier model for ~1/17 the token price
• Comes standard across any Genspark plan

Gen-1 is on par with frontier models like Opus 5 or Fable 5.1 giving you quality that used to cost $5 per million tokens used down to $.30 per million tokens. What this means in Genspark is when you use AI Slides in Standard mode, you get the same quality as if you were using Ultra, just with less credits consumed. We have an entire blog detailing why we built a model, how it was trained and evaluated, and everything else you need to know. 

Gen-1 is live and I’d love to hear what you all think, and to hear what else you would like to see with AI Slides?

Hi Everyone, Henry from Genspark here. I wanted to share with everyone how I’ve been using another Genspark agent: Agentbase. Agentbase lets you build a database you can talk to through an AI agent.

I collaborate with teams to gather data across multiple channels for help center documentation, customer feedback, and social media listening. The help center dashboard is where I spend the most time. It's where I go to see what needs updating whether it's adding a few FAQ’s, a new section, or an entirely new doc. I’ll walk through how this one works and how it connects to the other Agentbase dashboards. 

In this Agentbase project I have my dashboard with an agent I can talk to and ask it things like "show me everything still open," “what is needs to be done ASAP,” "what came in this week," and it queries, inserts, updates for me, and tells me what needs updating in the help center. It's connected to my Genspark account so I’m not manually adding everything, other agents in my workspace can update this dashboard for me.

How a request flows

Every entry Request moves through a simple lifecycle so nothing gets lost:

• Requested — something comes in that needs documenting.
• In Progress — me or my agents in GenTeam start working on it.
• Out for Review — colleagues double-check the doc.
• Done — the doc ships, the entry closes with a note on what changed.

Where it connects

The help center dashboard isn’t standalone. I have agents watching other dashboard and they add entries to mine based on what they see:

• Our social media dashboard surfaces what people are saying. When a theme is a documentation gap, it becomes an entry in my dashboard.
• When support answers the same question over and over because a doc is missing, it gets added to my dashboard.
• When we announce feature updates in GenTeam, my agents add the new features to my dashboard.

Why run it this way?

Documentation is easy to put off. Putting it in a dashboard with statuses everyone can see means I can tell anyone, anytime, exactly what shipped and why.

The challenging part: It's possible to have the same entry filed twice or for an entry to be added with no note/description. I’ve had entries land that I couldn’t act on because there was no information. Agentbase doesn’t dedupe entries or have rules for what's being entered so whatever is entered is there regardless if it's useful or not (unless prompted to do so).

I built all this inside Genspark. I set up the dashboards myself (with the Agentbases help) and the agents keep them updated. I'd love to hear how everyone else is tracking this kind of work!

Hi Everyone, Henry from Genspark here.

I help write and maintain our help center docs, and for a long time a brand-new one took 2-3 days: researching, drafting, fact-checking, rewriting, repeat. I had one basic AI assistant, but all the back-and-forth was on me.

I moved the whole thing into GenTeam and built a team of seven agents: an orchestrator, an information-gatherer, a researcher, a doc writer, a fact-checker, a proofreader, and a brand strategist. I prompt the orchestrator agent with a description and any specific details we want to add to the docs and it assigns tasks to each agent. Then each agent prompts the next, and a doc isn't called finished until it clears three things:

1. factually correct
2. makes sense and reads well
3. follows our existing guidelines, style, and format

The orchestrator and proofreaders keep bouncing the draft back until all three pass. The key part: this runs while I'm away, with no per-hop approval, so a doc goes through multiple revision cycles on its own and I come back to something close to done. The quality lives in the skills I configured with each agent.

This wasn't a plug-and-play solution. The skills needed to be refined and customized, with fair amount of re-prompting and refinement along the way.  The agents wanted to include unnecessary details to the docs and some claims still needed to be checked against the UI, which I haven't wired up yet. 

Net: It gets a doc about 90% of the way there. The last 10% is me.

On cost, since someone always asks: it isn't free. Every agent's turn consumes credits, and it scales with how many revision cycles a doc needs. I'm trading credits for time, and for a doc that used to eat 2-3 days, that's been worth it.

If you're setting workflows up, put the effort into the per-agent instructions and into deciding up front what each agent should do. I'd love to hear how everyone else is using GenTeam.

Hey everyone! I'm Henry, part of the team at Genspark. Credit questions come up quite often and I wanted to make a post to try and answer them here. If I miss something, ask below and I’ll be sure to answer you.

TLDR: On Plus/Pro, AI Chat and standard AI Image don’t consume credits (on standard mode). Tools that generate bigger outputs (Slides, Docs, Design, Video, Super Agent, etc) do. Failed generations aren’t charged. Credits reset each billing cycle and don’t roll over. Free plan AI Chat draws from a daily credit balance that refreshes everyday.

What’s free vs what costs credits? 

If Genspark is generating something – research, a deck, video, agent task, etc. that costs credits. Browsing, organizing, sharing, and hand-editing are free.

On Plus and Pro, AI Chat and AI Image are credit-free, with no monthly cap.

"I was just chatting and my credits disappeared. What happened?"

This is the #1 source of confusion, and I want to address this directly. AI Chat is credit-free on paid plans, but Super Agent is not even if you ask it the same question. Super Agent does web research, runs tools, and generates long outputs, which is all metered. If your credits dropped and are not sure why, this could be why

“Unlimited chat… but there’s a rate limit”

On Paid plans, AI chat and standard AI Image have no credit cost. If you pick a paid image model, you’ll see a confirmation box before it gets generated so there shouldn’t be any surprises. This is what unlimited means here. 

There are session rate limits. If you’re sending a ton of requests in a short window, it’ll pause you for a few hours and then reset. This is an anti-abuse measure and any normal use will not hit this limit even as a heavy user. 

How does the free plan work? 

Each chat/image draw from a daily credit balance that refreshes daily. If it runs out, your session will be paused until the credits are refreshed. 

What tasks cost the most? 

Credit consumption depends on the task at hand. Here are some rough tiers of this: 

Cheap: text tasks in Super Agent, short docs, basic spreadsheets

Medium: slide decks (scales with slide count), Code tasks, custom agent workflows

Expensive: Design (render heavy), video generation (especially long clips), large models with long context

Do I get charged when a generation fails? 

No, if a generation fails you shouldn’t lose any credits. If you believe you were charged for a failure, contact support and they’ll look into it.

Do unused credits rollover? 

No, credits refresh at the start of each billing cycle and unused ones don’t carry over. 

How I stretch my own credits (this can work for everyone):

• I draft my prompt with AI Chat first (it's free), then take it to the agent I need to perform the tasks. I see a lot of wasted credits come from re-runs caused by vague prompts 
• Be specific up front. “5 slide pitch deck on Q3 marketing results, exec audience, dark theme with charts” will take one generation. “Make me some slides about marketing” will require multiple rounds of editing after its generated
• If there is one slide that I don’t like. I specifically fix that one slide rather than re-generating a whole new slide deck. Costs way less credits this way.
• I match the model I use to the job at hand. For quick tasks or just need a simple output, I use the most efficient model.

 

These are the questions I see pop up the most. If I missed something, drop it below and I’ll answer you in the comments and update the post!

TL;DR: Know what you want to make? Use the dedicated agent. Not sure? Start with Super Agent or AI Chat and let it guide you.

Hey everyone! Henry from Genspark here.

I see a lot of people using Super Agent as their go-to for everything. But I wanted to share some tips that I've been using to level up the output quality, especially for specific tasks.

Genspark has dedicated AI agents for just about everything — images, video, slides, docs, music, voice, podcasts, design, code, spreadsheets, even full workflow automation. Whatever you're trying to make, there's an agent built specifically for it.

When you know what you want to create, go straight to the dedicated agent.

Here's the lineup for what I use:

Creative & Design: 

• AI image
• Genspark Design
• AI video

Documents & Presentations:

• AI Slides
• AI Docs
• AI Sheets

Technical:

• AI Developer
• AI Chat

 

Automation & Productivity:

• Genspark Claw
• SuperAgent

 

Why I go direct?

Each dedicated agent has been fine-tuned for its specific task — the prompting, the models, the output formatting, everything is optimized for that one thing. When you use a dedicated agent directly:

• Better output quality — the agent uses task-specific models and settings
• More control — you get the full set of options and customization for that tool
• Faster results — no routing overhead, goes straight to the right pipeline

When do I use Super Agent?

I use Super when:

• I’m exploring and not sure which tool fits my needs
• my task spans multiple agents (e.g., "research this topic AND make a slide deck about it")
• I want a conversational workflow where where I can iterate back and forth

My Takeaway

When I use Genspark's dedicated agents for specific tasks I get better, faster results than routing everything through Super Agent. Each agent is purpose-built for its thing, so the output quality, customization options, and speed are all noticeably better when you go direct.