
2026/6/25 · 10:35
Genspark: The $2.6B Super Agent That Turns "Vibe Working" Into a Credit Meter
Genspark sells an all-in-one AI workspace with demos, funding, and real phone calls; the catch is an opaque credit economy, uneven reliability, and users who say the meter moves faster than the work.
リサーチノート
The pitch
Genspark wants to sell you the fantasy that one AI workspace can replace your browser, your slide deck, your call center, and the boring parts of work all at once. In the company’s own language, it is an "all-in-one AI workspace" built around a Super Agent, AgentBase, and a promise that agents will handle execution instead of just generating text. The slogan in its own blog is "Less Control, More Tools." That sounds bold on a stage. It also sounds like a decent way to describe a machine that is about to start billing you while it thinks. 1 2
The money story is even louder than the product story. Genspark says it closed a $100 million Series B extension, bringing total Series B funding to $485 million at a $2.6 billion post-money valuation. The same release says the company added another $150 million in ARR in the first three months of 2026, on top of $100 million ARR since launch in April 2025. Investors clearly bought the narrative that this is not just a tool, but a category. 1
The demo story is strong enough to understand why people keep trying it. The company’s own blog says the Super Agent researched conference speakers in real time, drafted slides, made a live phone call, turned a prompt into a website, analyzed a spreadsheet with Jupyter, and pulled Gmail and Drive into the same flow. This is the kind of product demo that makes a room go quiet for 30 seconds, because it looks like the software is finally doing the work humans hate. 2
What the numbers actually say
| Signal | What the receipts say |
|---|---|
| Funding | Genspark says it closed a $100M Series B extension, bringing the round to $485M at a $2.6B post-money valuation. 1 |
| Growth | The same release says the company added another $150M ARR in the first three months of 2026, after reaching $100M ARR since launch. 1 |
| Hands-on performance | An affiliate-disclosed review gives Genspark a 7.3/10 rating and says phone calling worked on 39 of 47 calls, with 83% phone success and 87% research accuracy. 3 |
| Pricing | A third-party review lists Plus at $24.99/month for 10,000 credits and Pro at $249.99/month for 125,000 credits. 4 |
| User sentiment | Trustpilot shows 136 reviews and a 1.7/5 score, and its AI summary says people found the product low quality, clunky, error-prone, and backed by support that is "awful, non-existent, and unresponsive." 5 |
| Customer complaints | Reddit users say a Pro account "burned through it in a single day," support was "zero," and Claw drained about 20,000 credits in less than 36 hours. 6 7 8 |
The point of that table is not that Genspark is fake. It is not. The point is that the product can be genuinely useful and still be built around a meter that users do not trust. The review data shows real capability, but not enough predictability. The complaint data shows the same thing from the user side: the outputs may be fine, but nobody seems to know when the meter starts, how hard it runs, or who is supposed to answer when it breaks.
The reality check
The nicest possible read is that Genspark is an aggressive but capable agent platform that still has rough edges. The harsher read is that it is a very expensive way to discover that autonomy and cost control do not naturally like each other. One reviewer got 39 successful phone calls out of 47 and a decent research score. That is not nothing. But it is also not the kind of reliability you want when the product is charging by the credit and presenting itself as a substitute for a workflow stack. 3
The user complaints make the tension obvious. On Trustpilot, the recurring themes are fast credit burn, bad output, billing problems, and support that barely exists. On Reddit, the complaints are even less polite because they are less edited. One user says the product shut them down after a day and only gave them about two and a half decks. Another says support was "zero" and that reporting a problem got them nowhere. A third says Claw drained 20,000 credits in under 36 hours, then ate another 3,000 in a few hours after a top-up, which is about as far from "workflow efficiency" as you can get without setting money on fire on purpose. 5 6 7 8
That is the real problem. Not that it never works. Not that the demos are fake. It is that the product creates a trust gap every time it gets more autonomous. The more jobs it can do, the less obvious the cost becomes. The less obvious the cost becomes, the faster users start feeling like they are being ambushed by their own invoice.
The catch
Genspark’s own philosophy is the catch. "Less Control, More Tools" sounds great when the company is showing a live demo at a conference. It sounds very different when the same philosophy is applied to a paid plan that can vanish in a day, a Claw session that runs away with credits, or a support queue that never answers in time. The company is basically telling you that freedom is the point. Users are telling you that freedom is expensive.
A tool like this can genuinely help with one-off research, slides, call placement, or quick output when the stakes are low and the budget is loose. It becomes a much uglier bet when the work has to be repeatable, when spend needs to be forecasted, or when the people using it are not going to laugh off a surprise bill. If a product is supposed to reduce busywork, it should not create a second job called "watch the credits."
Verdict
Use Genspark if you want a flashy agent demo that can actually do a few real things and you are fine paying for the thrill ride. Skip it if you need predictable usage, clean support, reliable billing, or a team tool that will not turn every task into a scavenger hunt for the credit balance.
The funding round proves investors like the story. The reviews prove users are still paying for the rough draft. Right now, Genspark looks less like an AI workspace and more like a very expensive lesson in what happens when the meter has more personality than the agent.
参考ソース
- 1Genspark.ai Extends Series B to $485M at $2.6B Post-Money Valuation, Appoints Jamison Powell as Chief Revenue Officer
- 2Less Control, More Tools: How Genspark Built a Super Agent That Scales
- 3Genspark AI Review 2025: AI Agent Phone Calls & Workspace
- 4Genspark Review: Features, Pricing, and Alternatives in 2026
- 5Genspark Reviews 136
- 6Signed up for Genspark Pro… lasted one day? Seriously?
- 7Don't buy GENSPARK at any cost! Warning.
- 8Genspark Claw is a credit drainer

このコンテンツについて、さらに観点や背景を補足しましょう。