Wix.com reported its first-quarter 2026 results on May 13 and used the earnings release to confirm a significant strategic shift: the website-building company has built and deployed its own proprietary large language model, which is now the engine running under Wix Harmony, its hybrid vibe-coding and visual design platform.
The announcement lands as Wix increasingly leans on AI to differentiate from low-code rivals and as enterprise software vendors across the board reassess how heavily to depend on frontier model providers like OpenAI, Anthropic, and Google.
A proprietary LLM, not just a wrapper
In its shareholder letter and earnings commentary, Wix said its internal testing shows the proprietary model is meaningfully faster than alternatives, produces fewer errors, and delivers noticeably better outcomes for users building Harmony websites. The company described the move as a flywheel: by continuously fine-tuning on its own platform data and user feedback, Wix expects to widen the quality gap against generic third-party LLMs over time.
Wix also pointed to a more prosaic motivation — inference costs. By owning the model, the company said it can scale Harmony "with little to no reliance on third party LLMs," giving it direct control over a cost line that has become a major operating expense for AI-heavy SaaS businesses. Wix indicated this is the first in what it expects to be a broader portfolio of proprietary models across its product surface.
Q1 numbers: growth, buybacks, and AI traction
Financially, Wix reported revenue of $541.2 million, up 14% year over year, and bookings of $585 million, up 15% year over year. Total annual recurring revenue reached roughly $1.903 billion.
The company also highlighted unusually strong traction from newer AI products. Bookings from new user cohorts rose about 46% year over year, and Base44 — the AI app-building product Wix acquired and integrated last year — reached approximately $150 million in ARR as of May, according to the earnings materials.
In early April, Wix completed a $1.6 billion tender offer that retired nearly 30% of its outstanding shares, a sizable buyback that signals confidence in cash generation even as the company invests heavily in AI infrastructure.
Implications: SaaS rethinks the model layer
Wix's move sharpens a question many vertical SaaS leaders are asking: at what scale does training and serving a domain-specific model in-house become cheaper and better than calling out to a frontier LLM? Wix is betting that for the narrow, repetitive task of generating and refining websites, the answer is now — particularly when paired with proprietary user-feedback data the company already collects.
If the strategy works, expect more application-layer companies with deep vertical data to follow, eroding a slice of the API revenue that frontier model providers have come to rely on.
