Phrase

Where Phrase sits

Phrase is the enterprise end of the translation management market. The company is the union of PhraseApp (the developer-oriented localization tool) and Memsource (the CAT/TMS platform), now sold as one Phrase Localization Platform whose pieces — Phrase TMS, Phrase Strings, Phrase Language AI, Phrase Orchestrator, Phrase Studio, Phrase Portal, Phrase Custom AI — cover everything from string extraction to machine translation routing to multimedia localization. In 2025 Forrester named it a Leader in the analyst firm's first-ever Wave for translation management systems, and that positioning tells you who the buyer is: multi-vendor localization programs, LSP workflows, regulated content, companies with a localization manager and a budget.

It also tells you what happened to pricing. Phrase announced single-platform subscription pricing in January 2024 with three business tiers — Starter at $135/month, Team at $1,045/month, Business at $4,395/month, all billed annually. Over the following two years the ladder was dismantled in stages: Business went to "custom" on the public page by January 2025, Starter disappeared entirely between August and October 2025, and by mid-2026 the entry business plan (Team) listed around $1,245/month billed annually — roughly $15,000 a year — behind a "Get in touch" call to action. What remains self-serve at time of writing is a $525/month developer plan (software UI/UX) and a $27/month Freelancer plan that is a single TMS seat for individual translators. The old batch data I was handed listed entry pricing at $175; that shelf no longer exists.

What it actually does

The capacity table on phrase.com/pricing is the honest core of the product. At the Team tier (the entry business plan) you get: unlimited projects in both Phrase Strings and Phrase TMS, unlimited languages, 20 Phrase Strings seats with unlimited Phrase TMS seats, 1,200,000 managed words, 2,500,000 TMS processed words per year, 12,000,000 machine translation units (MTUs) per year, 25,000 AI units per year, three Orchestrator workflows, up to 20 Phrase Portal users, and one deployed custom AI model. Business roughly triples-to-quintuples those allowances (150 Strings seats, 3M managed words, 12M processed words, 50M MTUs, 100,000 AI units, three deployed models). Enterprise is custom across the board. All plans include 24/7 support and a 14-day trial of the Language Intelligence Platform.

Behind those numbers sits a serious toolchain. Translation memory with priority ordering and metadata prioritization, term bases with custom fields, pre-translation, automated and manual QA, a web CAT editor, in-context preview and in-context editing, business units, custom metadata fields, and cross-project TM search — all present from Team upward. The machine translation layer is where Phrase differentiates: MT Autoselect picks an engine per content, Phrase NextMT and Next GenMT are Phrase's own engines, MT Profiles (6 at Team, unlimited above) and MT Glossaries (10 at Team) let you tune output per domain, and Phrase Custom AI deploys your own trained model. Orchestrator adds automation with Auto Adapt and MT Optimize. Phrase Quality Performance Score gives you a measurable read on output quality.

The integration grid is deep and specific about which product each connector belongs to: GitHub/GitLab/Bitbucket for both Strings and TMS, Figma and Sketch for Strings, Unity for Strings, Jira, WordPress, Joomla, Drupal, Zendesk Guide and Zendesk Support, Google Drive/OneDrive/Box, Amazon S3/FTP/SFTP/SharePoint, content platforms including Contentstack, Kontent.ai, Sanity.io, and a list of heavier CMS and marketing connectors (Adobe Experience Manager, Salesforce surfaces, Sitecore, Optimizely, Contentful, Tridion Docs) that are add-ons or tier-gated. Phrase Studio covers multimedia localization as an add-on throughout.

My testing notes

The first observation is about the seat-versus-words split, which I think Phrase gets right for enterprises and wrong for small teams. Unlimited Phrase TMS seats at every tier means your LSP's translators can all work in the system without procurement gymnastics — a real advantage over per-seat TMS pricing. But the Strings seats (20 at Team) and the Portal user caps are where small teams hit walls, and the word allowances are annual, not monthly: 2.5M processed words per year at Team is fine for a mid-size product, but if your workflow re-translates churny content you can exhaust it. The unit separation is genuinely complex — managed words, processed words, MTUs, AIUs are four different meters — and I needed the comparison table to hold them all in my head at once.

Second, the MT stack behaves like an orchestrator rather than a single engine. MT Autoselect routing across engines, profiles and glossaries per domain, and the quality scoring layer mean a localization manager can treat machine translation as a tunable supply chain. In my testing the practical win is consistency controls: glossaries and profiles applied before human post-editing cut the correction passes I would otherwise run. The Custom AI (deployed models: 1 at Team, 3 at Business) is the escalation path for teams with parallel corpora.

Third, developer ergonomics. The CLI-oriented workflow — strings extraction, GitHub sync, in-context editing for web, Figma integration for design-side copy — is more mature than most TMS competitors, and the API supports headless automation. What I look for in this category is whether a release engineer can wire localization into CI without babysitting; Phrase's documentation and tooling pass that bar better than the enterprise-packaging suggests.

Where it falls short

The obvious limit is money. The effective self-serve floor for a product team moved from $135 to $525 a month, and the entry business plan sits near $15,000/year behind a sales conversation. If your localization needs are one app, three languages, and a designer who updates copy weekly, Phrase is not for you at these prices — the Freelancer plan at $27/month is a single TMS seat aimed at translators, not product teams.

Second, complexity. Four metered units, tiered connectors, add-on surfaces (Studio, several CMS integrations), and a product matrix of seven named components mean implementation is a project. Personalized onboarding, workflow review, and AI deployment are themselves add-ons at Team and Business — for that money I would expect them included.

Third, the public pricing page publishes capacities but not prices for the business tiers, which makes independent budgeting awkward. Everything I quoted above about dollar figures comes from third-party trackers and journalism, at time of writing, not from Phrase's own checkout.

Verdict

Phrase is the strongest enterprise TMS I have evaluated, and it has deliberately traded the small-team market for that position. It rated 3.9 in my rubric — 40% core functionality (the deepest CAT, MT-orchestration and QA stack in the category, offset by complexity), 20% value (excellent per-unit economics at enterprise volume, poor below it, and the self-serve ladder has been dismantled), 20% usability (capable but dense), 10% support and ecosystem (24/7 support on all plans, unusually deep integrations, Forrester Wave Leader), and 10% reliability signals (established vendor, mature platform). If you run a multi-vendor localization program with LSP workflows and a localization manager on staff, shortlist it. Everyone else has been priced out on purpose.

What I compare it against

The honest alternative set for Phrase is narrow. Smartling is the closest like-for-like — similar enterprise gravity, similar sales-led motion, similar word-based economics — and choosing between them tends to come down to which ecosystem you already sit in and which CAT editor your linguists prefer. MemoQ and Trados Studio serve the LSP-heavy crowd that wants on-premise or desktop-grade tooling. At the developer end, Lokalise and Crowdin cover the strings-and-repositories workflow at a fraction of the price, with far less MT orchestration and QA depth. And for teams whose content is mostly support conversations rather than product strings, Unbabel-style risk-routed translation services are a different answer entirely.

Where Phrase wins the comparison in my testing is the breadth of a single contract: one platform covering TMS, strings, MT orchestration, QA scoring, portal workflows for external vendors, and multimedia add-ons. Where it loses is precisely the same thing seen from the buyer's side — you are buying a platform whether you need all of it or not, and the tier allowances force you to estimate annual word flows before you have any data. My practical advice to teams evaluating this: run the 14-day trial against your real string base, measure your actual processed-word and MTU consumption over a sprint, and only then extrapolate to the annual allowances. The meters are published clearly enough that a trial gives you real numbers, which is more than most enterprise platforms offer.

One more point on the AI layer, since every TMS now claims AI. Phrase's AI story is deeper than a chat box bolted onto the editor: deployed custom models, MT Optimize and Auto Adapt in Orchestrator, AI Translation Agent, and per-plan AI unit budgets (25,000 AIUs at Team) show AI as metered infrastructure rather than a gimmick. The flip side is that AI consumption is a fourth meter on your bill, and AIUs are annual allowances, so a workflow that leans on AI rewriting heavily will need the Business tier or negotiated Enterprise volumes.

Sources

• https://phrase.com/pricing?currency=us
• https://phrase.com/pricing?tab=phrase-tms
• https://www.locize.com/blog/phrase-lokalise-price-changes-2026
• https://www.locize.com/compare/phrase-alternatives
• https://aiproductivity.ai/pricing/phrase
• https://phrase.com/product-launch-september-2026/