Base44 launches its own AI model to stand out in vibe coding
The Wix-owned vibe coding platform Base44 has started rolling out its own AI model, betting that owning the full stack will improve latency, cut inference costs, and eventually outperform frontier models. The move comes as AI coding startups race for long-term defensibility beyond third-party LLM wrappers.
Base44, acquired by Wix for $80 million roughly one year ago when the company was barely six months old, lets users create apps through natural language prompts. Its latest step — deploying a custom large language model trained on real user interactions — signals a broader industry shift. Players with strong brands are leaning into proprietary data and infrastructure rather than relying solely on external model providers.
Key Takeaways
- Base44 is rolling out its own AI model to power natural-language app creation on its platform.
- Founder Maor Shlomo says owning the model enables optimizations on latency, cost, and efficiency.
- The launch reflects growing debate over whether frontier models suit every use case — and whether wrapper startups are defensible.
- VC appetite for AI coding remains intense, with Chamath Palihapitiya's 8090 Labs closing a $135 million Series A the same week.
- Google is simultaneously expanding free AI features, underscoring fierce competition across the stack.
Why is Base44 building its own AI model?
The Bay Area-based company frames the rollout as a strategic bet on vertical integration. In coverage by TechCrunch, Shlomo said training and owning the model as part of the entire stack allows far more control over performance trade-offs.
Base44 also cited direct control over compute and inference spending, which it expects will produce a structurally stronger margin profile over time. That matters in vibe coding, where inference costs can weigh heavily on unit economics as platforms scale.
Does owning a model make vibe coding startups defensible?
Discussion in AI circles has intensified over two related questions: whether frontier models are best suited for all tasks, and whether businesses built atop someone else's models can endure. Base44's move speaks to both.
While the custom LLM is only just rolling out, the company hopes it will eventually outperform frontier alternatives as training continues on platform-specific data. Shlomo is betting that the engineering effort behind the model — referred to internally as Base1 — will cement Base44 as a vertically integrated vibe-coding player that owns distribution, data, and infrastructure simultaneously.
For more on how AI platforms are evolving beyond simple chatbot wrappers, see our Future Tech & AI Wonders coverage.
How heated is the AI coding startup race right now?
Investor enthusiasm shows little sign of cooling. On the same day Base44 announced its model rollout, Chamath Palihapitiya revealed that his AI coding startup, 8090 Labs, closed a $135 million Series A led by Salesforce Ventures. Palihapitiya also said he is stepping in as CEO, calling the moment comparable to the early social-media boom he witnessed as a Facebook executive.
8090 Labs targets corporate programming teams building production-quality software with enterprise controls like audit trails — but the funding wave underscores how crowded and well-capitalized the category has become. Rival Lovable reportedly reached $500 million in annual recurring revenue earlier in June, setting a high bar for growth expectations across the sector.
What else is shifting in the broader AI landscape?
Competition is not limited to coding tools. Google expanded Gemini's personalized AI image generation to eligible free users in the United States, allowing the chatbot to create images tailored to a user's interests. The move highlights how major platforms are bundling generative features at no cost to retain users — another pressure point for specialized startups seeking differentiation.
For Base44, the near-term test is whether its proprietary model delivers measurable gains in speed and cost before rivals close the gap with their own infrastructure bets. The vibe coding wars are no longer just about slick interfaces — they are increasingly about who controls the intelligence underneath.