Genspark
Last reviewed
Jun 4, 2026
Sources
18 citations
Review status
Source-backed
Revision
v1 · 2,435 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 4, 2026
Sources
18 citations
Review status
Source-backed
Revision
v1 · 2,435 words
Add missing citations, update stale details, or suggest a clearer explanation.
Genspark is an AI agent and productivity product developed by MainFunc, Inc., a startup based in Palo Alto, California, with an office in Singapore. It was founded in 2023 by two former Baidu executives, Eric Jing (Jing Kun) and Kay Zhu (Zhu Kaihua), and first launched to the public in June 2024 as an AI-powered search engine that generated custom result pages called "Sparkpages" instead of returning links. In April 2025 the company pivoted toward agentic AI, releasing the Genspark Super Agent, an autonomous general-purpose agent that coordinates multiple large language models and a library of tools to complete multi-step tasks. Genspark drew attention for unusually fast revenue growth and for a "Mixture-of-Agents" architecture that routes work across models from OpenAI, Anthropic, and Google. It reached unicorn status (a valuation above $1 billion) within roughly a year and a half of launch.
MainFunc was started in 2023 by Eric Jing and Kay Zhu, who had worked together at Baidu for about eleven years before leaving in October 2023. Jing, who goes by Eric Jing in English (his Chinese name is Jing Kun), had been a vice president at Baidu and the chief executive of Xiaodu Technology, Baidu's smart-device subsidiary that built voice-controlled hardware comparable to Amazon's Echo line. Earlier in his career he worked at Microsoft, where he was a research-and-development director on the Bing search team in Asia and is widely credited as a creator of Microsoft's "Xiaoice" conversational chatbot. Kay Zhu (Zhu Kaihua), the company's chief technology officer, had been a director-level architect associated with Google and the chief architect of Baidu's intelligent-assistant and smart-device efforts, and served as chief technology officer of Xiaodu. The two co-founded Xiaodu together before starting MainFunc. Wen Sang has been reported as the company's chief operating officer.
The company is incorporated as MainFunc, Inc. (sometimes written Mainfunc) and is headquartered in Palo Alto, with a significant presence in Singapore. Reporting has noted that a majority of its employees are Chinese, and the firm has positioned itself as a US-based company with cross-border roots.
Genspark launched publicly on June 18, 2024 as an "AI agent engine" for search. Rather than acting as a directory that points users to third-party websites, Genspark generated a synthesized, custom-built page for each query in real time. The company branded these generated pages "Sparkpages," describing them as a distillation of relevant information drawn from across the web and presented on a single, ad-free page. The product launched into a crowded field of generative search tools and was frequently compared to Perplexity and to AI overviews from Google. From the start, Genspark used several specialized models rather than a single model, assigning different query types to different components.
On April 2, 2025, Genspark released the Genspark Super Agent, a general-purpose autonomous agent that marked a strategic shift away from search and toward task execution. Instead of returning a page of information, the Super Agent was designed to plan and carry out multi-step jobs end to end: researching a topic, building a slide deck, assembling a spreadsheet, generating images or video, writing and deploying small web applications, and even placing phone calls on the user's behalf. The release positioned Genspark directly against other general agents that emerged in early 2025, including Manus and OpenAI's Deep Research, and against agent features from the major model labs. Over the course of 2025 the company de-emphasized standalone search entirely and rebuilt itself around an "AI Workspace" of productivity tools.
Genspark raised capital across three publicly reported rounds. The figures below are drawn from company announcements and from press reports by outlets including TechCrunch, Reuters (as relayed by SiliconANGLE), and the company's own press releases.
| Round | Date | Amount | Valuation | Lead / notable investors |
|---|---|---|---|---|
| Seed | June 2024 | $60 million | ~$260 million post-money | Lanchi Ventures (formerly BlueRun Ventures China) |
| Series A | February 2025 | $100 million | $530 million post-money | A group of US and Singapore-based investors |
| Series B | November 2025 | $275 million | $1.25 billion post-money | Emergence Capital (with SBI Investment, LG Technology Ventures, Pavilion Capital, Uphonest Capital, Lanchi Ventures) |
The June 2024 seed round of $60 million was unusually large for a company at launch and was led by the Singapore-based firm Lanchi Ventures, previously known as BlueRun Ventures China. TechCrunch and other outlets reported the seed valuation at roughly $260 million.
In February 2025, Reuters reported that Genspark had raised about $100 million in a Series A at a $530 million post-money valuation, roughly doubling the seed valuation in about eight months. The round was described as coming from a group of US and Singapore-based investors. (Some accounts label this round and its timing as March 2025 and cite a yuan-denominated valuation of about 3.85 billion RMB, which corresponds to the same dollar figure.) At the time of the Series A, Genspark reported around 2 million users, up from roughly 1 million monthly users in November 2024.
On November 20, 2025, the company announced a $275 million Series B at a $1.25 billion post-money valuation, led by Emergence Capital, an early backer of Salesforce. Named participants included SBI Investment, LG Technology Ventures, Pavilion Capital, and Uphonest Capital, alongside existing investors such as Lanchi Ventures. The Series B was announced together with the launch of the Genspark AI Workspace, and it pushed Genspark across the $1 billion valuation threshold, making it a unicorn roughly a year and a half after its public launch. Estimates of total funding raised range from about $335 million (summing the three disclosed rounds) to figures of $435 million reported by some outlets that count additional capital.
Genspark evolved from a single search product into a suite of agentic tools branded collectively as the AI Workspace. Key components include:
| Product | What it does |
|---|---|
| Super Agent | The flagship autonomous agent that plans and executes multi-step tasks, calling on sub-agents, models, and tools |
| Sparkpages | The original real-time generated search pages that synthesize web information into a single result |
| AI Slides | Generates editable, interactive presentation decks from a prompt |
| AI Sheets | Produces spreadsheets, including data tables enriched with images and visuals |
| AI Docs | Generates business-ready documents and long-form writing |
| AI Drive | Cloud storage for the files the agent creates, with versioning |
| AI Pods | Turns content into audio, including podcast-style narration |
| AI Browser | A dedicated agentic web browser with built-in Super Agent capabilities, ad blocking, and tool integrations |
| Call For Me | An agent that places real phone calls on the user's behalf to book reservations, schedule appointments, and handle errands |
The Super Agent is marketed as an "executive-level" agent that breaks a request into subtasks, dispatches each to a specialist sub-agent or model, executes the steps (including web searches, data analysis, image and video generation, and code), and assembles the results into a finished deliverable. Examples the company highlights include building slide decks, financial models, business documents, full-stack web applications, mobile apps, and even simple games from a natural-language prompt.
"Call For Me" is one of Genspark's most-discussed features. The agent dials businesses using a synthetic voice and conducts a natural conversation to complete tasks such as restaurant bookings or appointment scheduling, then returns a summary and recording of the call. In February 2026, Genspark and Twilio announced a partnership in which Twilio's programmable voice infrastructure powers the global calling capability behind the feature. Earlier coverage noted that the calling feature initially supported businesses in a limited set of countries, including the United States, Canada, and Japan.
Genspark released a dedicated AI Browser for Windows and Mac in 2025, positioning it as an agent-first alternative to conventional browsers. The browser embeds the Super Agent, blocks ads by default, supports an "Autopilot" mode for autonomous browsing, and connects to a large catalog of external tools and integrations (including through an MCP-style tool store). The company has framed the browser as part of its broader push to let an agent operate the web directly on behalf of the user.
Genspark's defining technical idea is what it calls a "Mixture-of-Agents" architecture. Rather than relying on one model, the system coordinates a set of large language models of different sizes and strengths together with a large collection of external tools, and a coordinating layer decides which model and which tools to use for a given step. Company descriptions of the Super Agent at its 2025 launch referred to roughly eight to nine differentiated LLMs (including models in the GPT-4o, Claude, and Gemini families, and in some descriptions DeepSeek), more than 80 tools or toolkits, and a set of proprietary datasets. By the time of the AI Workspace launch in late 2025, the company described the platform as drawing on more than 30 AI models, over 150 in-house tools, and more than 20 premium datasets, integrating with hundreds of external work applications. The "AI Chat" surface also exposes the Mixture-of-Agents idea directly to users by letting a query be answered by several models at once.
The approach is essentially a routing-and-orchestration strategy: by combining many third-party foundation models with its own tools and data, Genspark aims to outperform any single underlying model on real-world, multi-step tasks. The company has not fully disclosed the internal details of how its orchestration and model-selection logic works.
When it launched the Super Agent, Genspark promoted results on GAIA, a benchmark designed to measure how well general AI agents complete complex, multi-step real-world tasks. The company reported an overall GAIA score of 87.8%, which it said was ahead of Manus (reported at 86.5%) and well ahead of OpenAI's Deep Research (reported at 67.9%). Genspark also published per-level figures, citing about 87.8% on Level 1 (simpler tasks), about 72.7% on Level 2, and about 58.8% on Level 3 (the hardest tasks). These figures are the company's own self-reported results and should be read as vendor claims rather than independently audited measurements, though they were widely repeated in coverage of the launch, including by VentureBeat.
Genspark became a frequently cited example of how quickly an AI product could scale revenue in 2025. Multiple outlets and the company reported that the Super Agent reached roughly $10 million in annualized recurring revenue within about nine days of its April 2025 launch, about $36 million within roughly 45 days, and crossed an annualized run rate of about $50 million within roughly five months. In its November 2025 Series B announcement, the company said the platform was generating more than $500 million in annualized recurring revenue, a figure attributed in part to early-adopter and enterprise testing. Independent trackers have published a range of intermediate ARR estimates over the same period.
On the user side, Genspark reported passing 1 million monthly users in November 2024 and about 2 million users by the February 2025 Series A. Third-party trackers have cited figures on the order of 2 million monthly active users and roughly 100,000 paying seats, with team pricing reported around $30 per user per month. One earlier account of the company's first viral product cited about 10,000 paying customers at $99 per month; these third-party figures vary by source and date and should be treated as estimates.
Genspark's leadership has attributed the speed of its product output to an internal "vibe working" culture in which its own agents are used heavily to build and ship features. Chief executive Eric Jing summarized the company's pitch in the Series B announcement: "You state your intent, we deliver the finished work. This is a fundamental shift in how a billion people will work."
Genspark has been positioned by observers within the broader 2025 to 2026 wave of "general AI agents" that also includes Manus, Devin, and agent features from established labs. Commentators have highlighted its strong revenue-per-employee economics relative to its small headcount and its rapid pace of product launches. At the same time, some coverage has raised caution about the durability of headline ARR figures for fast-growing agent startups, and at least one 2025 security analysis of the AI Browser flagged concerns about agentic browsers being manipulated by malicious web pages. As with other products that orchestrate third-party foundation models, analysts have also noted the strategic question of how a model-agnostic agent layer sustains a moat as the underlying models improve.