Genspark

Genspark is an AI agent workspace built by Mainstream Inc., led by CEO Eric Jing, a former VP at Baidu who led the Xiaodu smart speaker division and earlier helped develop Bing for Asian markets at Microsoft. The platform launched in 2024 as an AI search engine producing "Sparkpages," structured single-page research overviews synthesized from live web data. In April 2025, Genspark pivoted hard to agentic AI, releasing Super Agent: a general-purpose autonomous system that takes a natural language request and works through it end-to-end, completing tasks rather than answering questions. By November 2025, backed by a $275M Series B at a $1.25B valuation from Emergence Capital, SBI Investment, and LG Technology Ventures, it had crossed $50M ARR in under five months. A further $385M extension in April 2026 brought valuation to $1.6B.

The core differentiator is a Mixture-of-Agents architecture: a central orchestrator coordinates 9+ distinct large language models, more than 80 in-house tools, and over 10 proprietary datasets. Claude handles planning and agentic reasoning; GPT-4-class models handle deep research and creative writing; Gemini handles multimodal understanding. The system can generate full slide decks in roughly four minutes, produce cited Sparkpages in under 15 seconds, build and deploy simple websites, place real voice phone calls for appointment booking, synthesize multi-day travel itineraries with live pricing, and create short video clips. It scored 87.8% on the GAIA benchmark at launch, outpacing Manus at 86%. A free tier offers 200 credits per day with unlimited basic search; paid plans start at $19.99/month (billed annually) for the Plus tier.

Genspark Super Agent at a glance, April 2026

As of April 2026, Genspark's primary product is Super Agent. The platform runs across three tiers: Free (200 credits/day), Plus at $19.99/month (annual), and Pro at $199/month (annual). The Mixture-of-Agents system coordinates 9+ foundation models dynamically based on task complexity and accuracy demands. Claude handles planning and reasoning; GPT-4-class models cover deep research; Gemini handles multimodal inputs.

Current capabilities include autonomous multi-step research with live web access, Sparkpage generation (cited pages combining text, video embeds, and tables), slide deck generation, website building, spreadsheet analysis, and AI phone call automation (83% completion rate in third-party testing). AI chat and standard image generation do not consume credits through December 2026. Video generation, complex agent runs, and AI calling do. Credits expire at month end with no rollover.

What Genspark Super Agent is actually good at

Genspark's clearest strength is research synthesis with visual output. Where Perplexity returns a clean text answer with citations, Genspark generates an interactive Sparkpage: a full structured document with embedded YouTube clips at relevant timestamps, comparison tables, images, and a chat interface for refining specific sections. For product comparisons and procurement research, this format saves real time versus opening twenty browser tabs.

The autonomous task execution is real. Third-party testing of the phone call feature recorded 83% completion across 47 test calls. David L., a management consultant in Boston, reported in September 2025: "Created a 20-page market analysis report that would have taken my junior analyst a full week." The GAIA benchmark score of 87.8% at launch outpaced both Manus and OpenAI Deep Research at that time. The backtracking capability, which lets the agent recover from failed sub-steps rather than stopping at an error, is a genuine architectural advantage over simpler linear pipelines. CEO Eric Jing framed the product philosophy directly: "We want our users to talk directly to the business outcome, instead of talking to a chatbot."

"The results are highly relevant and precise. The AI seems to understand the context of search terms exceptionally well.". Newton Karani, Product Hunt, 2025

"Created a 20-page market analysis report that would have taken my junior analyst a full week.". David L., management consultant, Boston, SpectrumAIReviews testing, September 2025

Where Genspark Super Agent breaks: the failure modes users keep hitting

Genspark's Trustpilot rating sits at 1.9 out of 5 as of early 2026. That signal deserves direct attention, because the complaints cluster into three specific patterns rather than vague dissatisfaction.

First, billing opacity. Multiple Trustpilot and Reddit reviewers describe non-consensual switches from monthly to annual billing, charges continuing after account cancellation or deletion, and no self-service cancellation button found in account settings during reviewer testing. This is a significant problem for any user who plans to trial the service and then stop.

Second, credit depletion confusion. Credits are consumed at rates ranging from 10 to 500 per task depending on complexity, with no pre-task estimate shown. One anonymous reviewer wrote: "I tried using the AI platform Genspark and the credit system is extremely restrictive. Almost every action consumes credits very quickly, which makes it difficult to properly test or use the platform." Credits also do not roll over: an Apple App Store reviewer noted, "The only thing I will not like is the fact that my credits do not roll over month to month." The salesdorado.com review found the platform "not a universal substitute" and flagged that service quality issues and billing opacity make it poorly suited to business-critical workflows.

Third, feature depth limits. Genspark's individual tools (slides, simple websites, video clips) each underperform their dedicated competitors. Slides are less polished than Canva output. Simple sites are less robust than Lovable builds. Video output lags Runway. The platform works best as a research and deliverable synthesis tool, not as a replacement for specialist creative software. Phone call reliability at 83% is good but not good enough for high-stakes appointment contexts.

Platform uptime during the 30-day SpectrumAIReviews test window was 94%, which means roughly one outage day per month. Support response times were flagged as slow across multiple review sources. These factors combine to make Genspark suitable as a supplementary productivity tool rather than mission-critical business infrastructure.

Genspark Super Agent vs. Perplexity vs. Manus

The three products that users consistently compare are architecturally distinct at a fundamental level.

Genspark Super Agent uses a Mixture-of-Agents loop with 9+ distinct LLMs, 80-150 tools, and a backtracking orchestrator. It executes multi-step tasks autonomously, including real API calls, voice calls, and multimedia generation. Output is typically a deliverable (a Sparkpage, a slide deck, a travel plan) rather than an answer. It scores 87.8% on GAIA. Pricing: Free tier / Plus at $19.99/month (annual).

Perplexity AI is a retrieval-augmented search engine at its core. It retrieves live web sources, synthesizes a text answer with inline citations, and returns it in a clean chat format. It recently added video generation for Pro users (Veo 3.1 integration), and Pro users get access to a "Labs" feature with limited agentic experiments. But Perplexity's fundamental architecture is search-plus-summary, optimized for accuracy and citation density. It does not execute multi-step tasks autonomously, cannot place phone calls, and does not generate rich multimedia Sparkpages. For academic research and factual verification, Perplexity's citation discipline is stronger. Pricing: $20/month Pro.

Manus AI (now integrated into Meta following a late 2025 acquisition, new signups paused) used a "virtual computer" architecture: sub-agents browse the web by mimicking human browser behavior, execute code in 20+ languages, and manage file systems. Its two core models (primarily Claude-based) handle planning and execution. Manus was oriented toward technical power users and developer workflows, with a waitlist and a technical interface. Genspark differs from Manus by routing across 9+ distinct LLMs rather than 2 core models, supporting 80+ more tools, offering a consumer-accessible zero-setup UI, and adding native voice/phone automation that Manus lacked. Genspark's GAIA score of 87.8% edged Manus's 86% at the April 2025 launch window.

The practical choice between Genspark and Perplexity comes down to output format: if you want a cited text answer, use Perplexity. If you want a structured research document, a slide deck, or an autonomous task completed end-to-end, Genspark's architecture is built for that job.

Is the paid tier worth it?

The free tier gives 200 credits per day, which is meaningful for light research use. Basic chat and image generation currently don't consume credits, so casual users can run a reasonable number of Sparkpages and standard queries without hitting limits. The free plan is a legitimate evaluation tier, not a funnel-only token offering.

The Plus tier at $19.99/month (annual) or $24.99/month provides 10,000 credits and 50GB storage. For a solo operator or freelancer using the platform for weekly research reports, competitive comparisons, or slide deck generation, 10,000 credits/month is workable given that standard research tasks consume 10-100 credits and complex agent tasks consume 100-500. The math holds if the use case is moderate-frequency research, not daily heavy automation.

The Pro tier at $199/month (annual) or $249.99/month with 125,000 credits and 1TB storage targets agencies and high-volume teams. At that price point, the comparison to running a dedicated research analyst or multiple specialist tools becomes relevant. The caution here is the billing opacity issue: the Trustpilot complaints about annual billing switches and cancellation difficulty suggest reading the terms of service carefully before committing to a paid subscription, particularly annual billing.

Credits do not roll over. If your use is front-loaded in a given month, unused credits at month end are lost. This is a notable structural friction versus flat-rate subscription tools.

Best use cases, and when to skip it

Genspark earns its place in workflows centered on research synthesis, structured document creation, and moderately autonomous task execution. Solo business owners comparing software procurement options get an interactive cited Sparkpage in 15 seconds. Consultants and analysts can generate multi-page research reports and slide decks in minutes. Trip planners get structured itineraries with live booking links. Teams that previously juggled browser tabs across review sites, Google Docs, and presentation software can consolidate several of those steps through Genspark's deliverable-first output model.

Skip it when you need guaranteed billing reliability, when mobile is your primary interface (Sparkpages are heavy for small screens), when you need deep domain accuracy for academic or legal research, or when a specific deliverable (a polished design, a robust web app) really needs a specialist tool. The 1.9/5 Trustpilot rating is a real risk flag for enterprise or business-critical deployments. If your primary need is fast, accurate web research with clean citations, Perplexity Pro at $20/month is a more predictable option with a stronger trust record.

Getting started with Genspark Super Agent

Genspark requires only an email signup, no waitlist. Web is the primary entry point; iOS and Android apps are available, though Sparkpages render better on desktop. The free 200 daily credits are enough to test core features: run a Sparkpage query, generate a slide deck, and use the chat interface before committing to a paid plan. Test a high-value use case first rather than experimenting broadly, since credits deplete quickly on agent-heavy tasks. Before subscribing to Plus or Pro, verify the cancellation process and billing terms carefully; the Trustpilot record on billing disputes warrants this step.