How to choose a TMS in 2026: pricing models explained, one project priced at nine vendors

Choosing a TMS (translation management system) comes down to two questions: does it fit how your team ships text, and does its pricing model fit how your project grows. The second question is harder than it looks, because vendors meter different things: words in all languages, words in the target languages, processed words, source keys, strings or seats. The same project can cost about €40 a month with one vendor and more than $500 with another. To make that concrete, we priced one reference project at nine vendors, Locize included, from their public pricing pages on October 13, 2026.

We are one of the vendors in the table, so the method is spelled out below, every price comes from the vendor's own pricing page, and where Locize is not the cheapest, the table says so.

Start with the workflow, not the price

A cheap TMS that does not fit your release process costs more in engineering time than any plan difference. Settle these first, then compare prices among the tools that pass:

• Integration with your i18n library and CI. Does it read and write your file format, sync from CI without a manual export, and report new keys from your app (with i18next, saveMissing)?
• How translations reach users. Files committed to your repo and shipped with the next deploy, or delivered at runtime from a CDN so text changes go live without a release? Both have their place.
• Review and quality. A real review step with roles, not only a "verified" flag; automated checks; and for AI translations, a way to see which ones deserve a look.
• AI and machine translation. Included, metered per word or token, or run on your own provider key.
• Who needs access, and whether those roles are billed.
• Exit. Export in your formats, an API that returns everything, no lock-in through proprietary placeholders.

The six ways TMS vendors meter

Transifex currently meters AI-translated words only, and publishes no rate.

Four of the nine combine a volume meter with a seat or role rule on the plans we priced. Read both, and read the role definitions: whether a reviewer counts as a seat can move a team up a plan.

The reference project

That makes 40,000 / 90,000 / 160,000 words across all languages, 30,000 / 80,000 / 150,000 words in the target languages, and 2,000 keys at every size. For each vendor we took the cheapest public plan that meets every need, and added public top-ups where a plan offers them.

What the same project costs

List prices on October 13, 2026, per month, without taxes and without AI or machine translation usage. First value: monthly billing; second: annual billing divided by 12, rounded.

Locize bills monthly only, so it has no annual column. SimpleLocalize's pricing page shows euros; Tolgee also shows euro prices (€69 / €107 / €179 monthly).

What decided each row:

• SimpleLocalize: Team (4,000 keys, up to 25 people) is the first plan for 6 people; Developer fits everything except people (5). Large language counts cost nothing extra. AI translation runs on your own LLM key; Google and DeepL characters are included up to a quota.
• Localazy: keys are the only meter, so all sizes cost the same. Automations and DeepL start at Autopilot ($94 / $78), branching and language permissions at Business ($210 / $175).
• Locize: Growth is the first plan with unlimited users and the review workflow; extra words come in blocks of 10,000 for $5 a month. L needs Professional because Growth covers 10 languages, source included. Growth adds 300,000 AI tokens with every monthly invoice, and unused tokens carry over up to a limit; your own OpenAI, Gemini or Mistral key works from Growth.
• POEditor: strings count each key plus each translation, so the price follows the language count. S sits exactly at the Plus limit. There is no official CLI; CI sync goes through the API or the Git integrations.
• Tolgee: the price follows hosted words in tiers; users are unlimited on paid plans. L at 160,000 words sits just above the 150,000 tier: if your translations are a little shorter than the source, it drops to $149 ($124 on annual billing).
• Crowdin: hosted words count target languages only. Team is the first plan where developers and the product manager get their own accounts with upload rights (5 managers included); translators and proofreaders are free. AI and machine translation run on your own key or a prepaid balance.
• Lokalise: bringing in the existing project at size M (about 90,000 processed words) already exceeds Explorer's yearly 60,000, and 15 targets exceed Explorer's 10 languages, so M and L need Growth. Users with only translator or reviewer permissions have free seats. On monthly billing the processed-word allowance is per month (Explorer 5,000, Growth 20,000), so the initial import needs add-ons (price not public) or several months.
• Phrase: the cheapest plan open to a software company is Software UI/UX at $525 a month with annual billing (quarterly billing is available on request; monthly is not offered), with 15 seats in which every user counts. On that plan, review is a verified flag; a formal review stage, branching and SSO need Business, which is sold through sales.
• Transifex: the pricing page shows one AI-first plan billed on AI-translated words, with unlimited seats and hosted content, and no public rate.

What the table shows

• For few languages, the meter matters less than the plan floor. At 3 target languages five vendors cost between about €40 and $80 per month; the expensive ones are expensive because their entry plan for a team with several uploaders or seats starts high. Within that group the difference is a few dozen per month, so the workflow fit decides more than the price.
• For many languages, key-metered plans win on price. SimpleLocalize and Localazy cost the same at 15 languages as at 3. Plans metered by words or strings (Locize, Tolgee, Crowdin, POEditor) step up as languages are added; Phrase's entry plan includes enough words for all three sizes. Among the word- and string-metered plans, Locize had the lowest monthly price at every size; at L, Tolgee's annual price comes within a few dollars if translations run a little shorter than the source.
• Seats decide more than the seat price. What changes the price is which roles are billed and where people caps sit: every user (Phrase), managers only (Crowdin), advanced seats only (Lokalise), a people cap per plan (SimpleLocalize, Locize below Growth) or nobody (Locize from Growth, Tolgee, Localazy, Transifex, and POEditor contributors from Start).
• Processed words follow change, stored words follow size. A project that rewrites its source copy every sprint uses more under a processed-word meter than under a stored-word meter with the same nominal allowance.
• Billing terms differ. Annual billing saves 10 to 17 percent on the plans in the table; one vendor offers no monthly billing at all; Locize offers monthly billing only.

When the project grows

A snapshot hides the steps. The same app a year later, with 5,000 keys and 25,000 source words (same 6 people), on the same rules, monthly billing:

• Key-metered plans move in steps at key limits: Localazy from $73 to $125, SimpleLocalize from €39 to €49 with two key add-ons.
• Word-metered plans grow with the words: Locize from $49 to $69 at S and from $129 to $249 at L, Tolgee from $79 to $119 and from $199 to $399, Crowdin unchanged at S and from $208 to $244 at L.
• The order changes: at S, Tolgee moves ahead of Localazy and POEditor drops behind both; at L, Crowdin moves ahead of Locize. SimpleLocalize stays the lowest list price in both.
• Lokalise at L outgrows the yearly processed words of Growth. The table prices the grown project as if it were brought in at once; growing inside Lokalise, year two at S uses about 60,000 processed words, exactly Explorer's allowance.
• At a limit: in the grown project, Tolgee at S, Localazy and SimpleLocalize sit exactly at a tier or key limit.

What the list price leaves out

• AI and machine translation usage. The vendors handle it in four ways: a monthly quota included in the plan, credits or prepaid balances, per-word rates, or your own provider key. A first full pass over 15 languages is the biggest single usage event; price it separately.
• Top-ups. Some vendors sell extra words, keys or seats but do not publish the price, so a project just above a limit can only be priced by asking.
• Features behind the next plan. Branching, SSO, a formal review stage, glossary, translation memory, QA checks and automation often start one or two plans higher. Check the plan you priced, not the product page.
• Tier edges. Several projects in the table sit at a limit: POEditor at S, Crowdin at L, Tolgee just above a tier at L, Locize at M (9 of Growth's 10 languages). A few more keys, words or languages move them to the next plan or add-on step (Locize at M: two more languages).
• Migration and engineering time. A TMS that does not fit your file format or CI costs more in hours than in fees.

How to price your own project

1. Count keys, source words, target languages, and the people who need access by role (developers, product, translators, reviewers).
2. Write down how the project grows: more keys, more languages, or more frequent copy changes.
3. For each shortlisted vendor, find the meter and the role rules on the pricing page and in its help center, then pick the cheapest plan that meets every need, not the cheapest plan.
4. Add the usage you will create: the first AI or machine translation pass and the monthly change rate.
5. Repeat for the size you expect in a year. The ranking often flips.

Which pricing model fits which team

• Many languages and a stable key set: key-metered plans (Localazy, SimpleLocalize) stay flat as languages grow, for any team size on Localazy and up to 25 people on SimpleLocalize Team. Watch the key limits, which come in steps.
• More people than developers touch the text: plans without seat fees (Locize from Growth, Tolgee on paid plans, Localazy, POEditor from Start, Transifex) keep the team out of the price; then the volume meter decides. Watch the word or key tiers as the project grows.
• Stable content, few managers and many translators or reviewers: a processed-word plan (Lokalise, with free translator and reviewer seats) charges content when it is created, imported or changed, not while it is stored.
• A large program with procurement, SSO and managed translation services: the higher plans (Phrase Business, Crowdin Enterprise with Team+ and Business, Lokalise Advanced and Enterprise) are bought through sales or a demo, some with a public starting price; compare them on the contract, not the list price.

Where Locize fits

Within the low-cost group of the table, price does not decide; what the tool does with your translations does. Locize is a good fit when one of these matters to you:

• AI translations that hold up, with a sign-off. AI translation uses your glossary, style guide and translation memory as context; each AI translation gets a confidence score, with review routing on, the doubtful ones go to review by themselves, an AI review checks the translations you already have, and the provenance export records per string who approved what: the documented human review that the EU AI Act's Article 50(4) exemption relies on.
• Coding agents write your strings. Through the MCP server (28 tools) an agent reports keys, reads the confidence scores and revises; its writes are recorded as agent writes, so the export does not count them as a person's review. A real run shows the loop.
• One product, many customers with their own wording. Tenants inherit a shared base and override only what a customer changes, instead of one copied project per customer; Growth includes 5.

Two things to check before choosing Locize: your keys and languages (for a stable key set in many languages, a key-metered plan costs less, as the tables show, and with the larger project at L, Crowdin's target-word meter comes in just under Locize), and your platforms (web apps, React Native and other runtimes that load JSON get text changes without a release; compiled iOS or Android string files do not, because Locize has no over-the-air SDK). Human translation comes through connected translation order services.

To check your own numbers, the pricing page shows words, languages and users per plan. Every account starts with a 14-day full-feature trial without a credit card, enough to connect a real i18next app and try AI translation with confidence scores, review and the MCP server on your own strings before you compare prices. Afterwards you pick a plan to keep the project, the free plan (2,000 words, 2 languages, 1 user) included. The comparisons cover the workflow side vendor by vendor.

Method and disclosure

• Sources: each vendor's own pricing page, plus its help center or documentation for what is metered and who counts as a seat, read on October 13, 2026. No third-party price lists. Where a page converts currencies by location, we used the vendor's base currency. Prices change; if you find a difference, tell us at support@locize.com and we will update the table.
• Rules: cheapest public plan that meets every need of the reference project; public top-ups only; the existing source and translations are imported in year one; no discounts, no taxes, no AI or machine translation usage. Where a plan meets a need only partly, the table says so (Phrase: basic review only).
• Approximations: words per target language are assumed equal to the source (real translations are longer or shorter); i18next plural keys are counted as keys like any other, although vendors may count plural forms differently.
• Conflict of interest: Locize is one of the nine vendors. We applied the same rules to our own plans, from the plan limits in our pricing code.

Frequently asked questions