Upwork is the largest general freelance marketplace on the internet, and for a huge range of work it earns that spot. If you need a defined project done — a landing page, a data pipeline, a contract mobile app — and you want a deep pool of candidates you can compare on price and reviews within a day, Upwork is a genuinely sensible place to start. This review is about where that model fits, and where hiring an AI-native engineer is a different problem.

Upwork is the product of two of the oldest names in online work. Elance launched in 1999 and oDesk in 2003; the two merged in December 2013 to form Elance-oDesk, then rebranded as Upwork in 2015. The company went public on the Nasdaq under the ticker UPWK on October 3, 2018, and is headquartered across Palo Alto and San Francisco. It's a real, public, profitable company — not a startup experiment. The scale is the whole point. In full-year 2025, Upwork reported record revenue of $787.8 million (up 2% year over year), gross services volume of roughly $4.0 billion flowing through the platform, and GAAP net income of $115.4 million. It ended the year with about 785,000 active clients. When people say "just post it on Upwork," this is the machine they're pointing at.

Upwork is an open two-sided marketplace. A client posts a job with a description and a pay range; freelancers spend "Connects" (tokens that cost about $0.15 each) to submit proposals, and the client picks from whoever applied — or invites specific people directly. Contracts can be hourly or fixed-price, and Upwork handles the payments, escrow, and reviews in between.

The fees are transparent, which is more than a lot of competitors can say. Since May 2025, freelancers pay a variable service fee of 0–15% per contract, set when they submit a proposal, with most landing around 10%. Clients pay a marketplace fee of roughly 5% on the Basic plan (about 3% for eligible U.S. bank transfers), plus a small contract-initiation fee of $0.99–$14.99 on new engagements. Upwork's overall take rate reached 19% in 2025, up from 18.1% the year before. You know what a transaction costs before you make it.

Selection and speed. With hundreds of thousands of active clients and billions in annual GSV, the supply side is enormous. For common skills you can have real proposals within hours. Few channels match that time-to-first-candidate. Transparent, self-serve economics. Escrow, milestone payments, hourly work-diary tracking, and published fees mean you can run a small engagement end to end without a sales call or a contract negotiation. For a one-off project, that friction-free flow is a real advantage. Reviews and work history. Every freelancer carries a public track record — job success scores, earnings, client feedback. For discrete, well-scoped work, that history is a useful signal you don't get from a résumé alone. Global reach and flexibility. You can hire across 180-plus countries, scale a freelancer up or down by the hour, and close the contract when the work is done. For variable or short-term needs, that optionality is exactly right.

It's a project marketplace, and it's optimized for that. The model shines when the work is defined and comparable. When you're hiring for open-ended engineering — someone who owns a system for months and makes judgment calls no ticket spells out — you're using a bidding platform for a role it wasn't built around. You do the evaluation. A job success score tells you a freelancer delivered past contracts well. It doesn't tell you they can architect against a moving AI stack. Screening for that — the interviews, the take-homes, the reference checks — is on you. You're choosing from who applied. A marketplace surfaces the people actively bidding for work this week. That's a feature for availability and a limit for reach: the strongest engineers in a given specialty often aren't scanning open job posts at all.

Here's the thing about an open marketplace: the best engineers usually aren't on it. Not Upwork, not any of them. The AI-native engineer you actually want is already working — booked on a contract somewhere else, heads-down on something hard, not buying Connects to bid on your job post. The people bidding are the people looking. The people you want aren't looking. Nextdev reaches the top 1% of AI engineers — the ones who aren't looking. We go find them, not wait for them to apply. Then we prove them. We give each a real problem to build and watch how they actually work — so you know they're AI-native because you've seen it, not because a job success score or a profile said so. That's the part a marketplace hands back to you: verifying the person can really do the work. We do it before you ever meet them. And when you want one, we employ them for you. One named engineer, on our payroll — the contract, the compliance, the paperwork never touch your desk. On Upwork you're the one running the engagement and carrying the risk. Here you're not. You're not picking from who signed up. You're getting the engineer who never would have.

Upwork is the right tool for a lot of work, and it's honest about what it is: a fast, transparent, deep marketplace for defined projects. If your job is scoped and comparable, start there — it's hard to beat on selection and speed. If your job is to hire one AI-native engineer who owns real work and passes real scrutiny, that's a different problem, and it's the one Nextdev is built for.

Sources