Freelance Platform Fee Comparison 2026

By Maya Ellison · Updated 2026-09-25

Direct answer: freelance platforms charge the freelancer between 0% and 20% of everything you earn — Upwork a variable 0–15% per contract, Fiverr a flat 20%, PeoplePerHour a tiered 20% → 7.5% → 3.5% by buyer lifetime, Contra 0%, Toptal no published commission (the markup sits on the client's invoice), and a direct client only card processing at about 2.9% + $0.30. On the same $1,000 project that is roughly $970 direct, $850–900 on Upwork, and $800 on Fiverr before withdrawal or conversion. A fee is worth paying while the platform is still your cheapest source of clients — and not a month longer. Below: the structures as published, a side-by-side table, and the arithmetic of one project across four models.

The 2026 fee landscape

Platforms justify their cut with client access, escrow, and dispute handling — but the percentages are material, and two structures changed recently. Everything below is as published by each platform; confirm on the platform's own fee page before you quote, because these schedules move without much ceremony.

As published: Upwork service fees, How Fiverr works and its payment terms, PeoplePerHour commission fees, Contra pricing. Check current terms before you build a rate on any of them.

The hidden costs beyond the fee

The headline percentage is only the first deduction. The rest stack in an order you almost never see on one screen:

• Proposal costs: Upwork Connects at $0.15 each turn business development into a marketing budget paid before you have earned anything. Burning 80 Connects a month on bids is $12 — small in isolation, real across a year of silence, and boosted bids cost more on top.
• The clearing period: Fiverr holds a new order 14 days before withdrawal. No fee attaches to that wait, but it is two weeks of working capital you finance — which matters the month you have bills and no payout.
• Withdrawal and conversion: $1–3 per transfer on Fiverr's payout methods, while PayPal or Payoneer currency conversion commonly costs another 2-4% on top.
• Availability pressure: Fiverr's algorithm rewards fast responses, which quietly prices in being always on — an unpriced cost that lands on your evenings.
• Exclusivity clauses: platform terms commonly bar taking a platform client off-platform for 12–24 months. Violating it risks a ban and forfeited earnings, so plan around the window instead of testing it.
• Buyer-side fees: Fiverr buyers pay 5.5% on top of your price and Upwork clients pay 3–10% plus a contract initiation fee, so clients experience your price as higher than the number you wrote.

Worked example: one $1,000 project, four models

Formula: net payout = project price − platform commission − withdrawal and conversion costs

Same price, same work, four ways of getting paid — all before the tax set-aside you owe on whichever row you land in:

• Direct client, card payment: $1,000 − (2.9% × $1,000 + $0.30) = $1,000 − $29.30 = $970.70.
• Upwork at 10% (mid-range of the published 0–15%): $1,000 − $100 = $900. At the 15% ceiling the same project pays $850, and eight Connects on the winning bid cost a further $1.20.
• Fiverr at 20%: $1,000 − $200 = $800. Withdraw to a US bank at $1 and you keep $799; convert currency at 3% first and it is about $775.
• PeoplePerHour, first invoice with a new buyer: the first £250 at 20% (£50) plus the remaining £750 at 7.5% (£56.25) = £893.75 kept — an effective 10.6%, because only the first slice pays the top tier.

The spread is the point: $170.70 separates the direct client from Fiverr on identical work — 17% of the project — and over twenty such projects that is $3,414 a year. Scale it the way a platform's finance team does: on $60,000 of Fiverr work, the flat 20% is $12,000, a marketing budget's worth of margin leaving your business whether or not the platform did anything extra that year.

Direct is cheaper — but not free

Going direct costs you client acquisition (sales time, maybe ads), invoicing and payment processing (~3%), and non-payment risk that escrow used to absorb. The honest comparison is fee versus acquisition cost: a 10% platform cut on a $1,000 project is $100, while five hours of outreach at a $75 effective rate is $375 of unbilled time — on those numbers the platform is still the cheaper channel. It stays cheaper only until your own cost to win a client falls below the fee. That is the crossover to measure with our client acquisition cost calculator: use platforms to build a reputation and land the first clients, compare what each channel really pays per hour, and move relationships direct when the maths says so — inside the platform's terms, never around them. Price the direct rate with the rate calculator, and see the full net-per-hour treatment in our side-hustle economics guide.

Which model fits this stage

Fees are one input, and the right answer changes as the business does:

1. No reputation yet: pay the fee. A marketplace supplying escrow, dispute handling, and a stream of strangers is cheaper than building all three yourself. Price the fee into your rate instead of apologising for it.
2. Some reputation, thin pipeline: stay, but instrument it. Track what each platform actually pays per hour after every deduction, and judge it against your direct clients with the same yardstick — the number, not the brand.
3. Repeat clients you could serve direct: graduate carefully. Wait for any exclusivity window to expire, use the platform's own permitted routes where they exist, and remember that direct means you now chase invoices and absorb disputes.
4. Steady direct work: drop to a listening post. Keep the profile alive at no marginal cost, decline anything that dilutes your rate, and spend the saved fee on the channel that produced your last three clients.

The failure mode is not paying a fee — it is paying one forever by default, on work that stopped needing the platform years ago.

Proposals or gig storefront: which buyer intent fits you

Upwork and Fiverr do not just charge differently — they sell to different buyer intent, and picking the wrong one costs more than the fee gap. The choice logic runs through three questions, all answerable from the fee structures above:

1. Does the buyer arrive with a defined project? Proposal marketplaces (Upwork's model: bid with Connects at $0.15 each, variable 0–15% per contract) fit buyers posting a scope and comparing bids — custom development, ongoing marketing, consulting. Gig storefronts (Fiverr's model: flat 20% including tips, 14-day clearing for new sellers) fit buyers browsing fixed packages — a logo, a voiceover, a quick edit. Match the channel to how your buyers already shop, not to which fee looks smaller.
2. How much unpaid selling can the margin carry? Proposals cost Connects before you earn anything — the $12-a-month bidding habit in the hidden-costs section — while storefronts cost margin on every order with no bidding outlay. If your win rate on cold bids is low, the storefront's 20% can be cheaper than months of paid proposals that convert to nothing; if you win bids efficiently, the 0–15% band plus modest Connects spend wins.
3. Where does the relationship go next? Repeat-buyer economics diverge: PeoplePerHour's tier (20% to 7.5% to 3.5%) and Upwork's per-contract lock-in reward staying where relationships compound, while Contra's 0% and the direct client's ~2.9% + $0.30 reward graduating proven relationships off-platform inside the terms. Judge each channel with the same net-per-hour yardstick from the worked example — $970 direct, $850–900 Upwork, $800 Fiverr — via our client acquisition cost calculator.

The first 90 days: a beginner earnings strategy

New sellers overpay fees in exactly one way: pricing as if the fee were temporary. It is not — so build it into the first quote and minimise what you can control around it:

1. Price the fee into the rate from day one. A $1,000 project keeps roughly $800 on Fiverr and $850–900 on Upwork before withdrawals, so a beginner needing $900 net must list at about $1,125 on Fiverr, not $1,000. Run the target through the rate calculator with the commission subtracted first — the arithmetic is in the worked example above.
2. Cap the bidding budget. At $0.15 per Connect, 80 blind bids a month is $12 plus boosted-bid extras — real money against zero revenue. Bid only where the scope fits work you can already show, and treat every Connect as spend from the marketing budget the $12,000-a-year illustration warns about.
3. Chase the tier drop, not the volume. On PeoplePerHour-style tiers the first £250 with a buyer pays 20% and everything after pays 7.5% then 3.5% — so one relationship pushed past its top tier beats three new ones stuck in it. Fiverr's flat 20% never drops, which is the signal to move proven buyers toward repeat packages or, inside the terms, toward direct work per the stage model below.
4. Withdraw deliberately. Batch payouts instead of dripping them: each Fiverr withdrawal costs $1–3, and currency conversion commonly adds another 2-4%. Fewer, larger transfers plus the 14-day clearing plan keep the fee stack to the commission row of the table rather than every row at once. The full net-per-hour treatment is in our side-hustle economics guide.