V0: Mini, Pro, and Max Model Tiers
V0V0 reintroduced model selection with three distinct tiers — V0 Mini, V0 Pro, and V0 Max — giving users control over the speed-intelligence tradeoff on a per-session basis. Mini is optimized for speed and cost efficiency, Pro targets superior performance for typical development tasks, and Max is designed for maximum intelligence on the most complex problems. The composite model architecture behind each tier combines RAG, frontier LLMs, and a custom AutoFix streaming layer.
Sources & Mentions
1 external resource covering this update
Model Choice Returns to V0
V0 brought back model selection with a new three-tier lineup: V0 Mini, V0 Pro, and V0 Max. The reintroduction gives developers direct control over the capability-cost tradeoff on a per-session basis — a feature that had been removed when V0 launched its Agent mode and is now back in a more mature form.
The Three Tiers
V0 Mini is the fastest and most cost-efficient option, suited for tasks with narrow scope — simple component generation, text edits, minor bug fixes, and quick iterations where speed matters more than depth.
V0 Pro delivers superior performance for the majority of development tasks. This is the general-purpose tier for building features, refactoring code, and working with moderately complex codebases.
V0 Max applies maximum intelligence to the hardest problems — complex architectural decisions, multi-file refactors, and generation tasks where output quality is the primary concern and cost is secondary.
The Composite Architecture Behind the Models
Vercel published a detailed breakdown of the composite model architecture powering the V0 model family. Each tier combines three specialized components:
- Retrieval-Augmented Generation (RAG) — specialized knowledge about frameworks, component libraries, and coding patterns is retrieved and injected at generation time
- Frontier LLM reasoning — the core generation draws on state-of-the-art large language models, with the specific underlying model varying by tier
- AutoFix streaming model — a custom post-processing model monitors the output stream in real time, catching errors and best-practice violations mid-generation before they reach the user
This architecture allows Vercel to upgrade the underlying base model as frontier models improve without rebuilding the specialized layers around it.
Pricing Implications
Model selection has direct credit cost implications — V0 Mini consumes fewer credits per generation than V0 Max. This gives developers an explicit lever to manage usage costs on tasks where top-tier reasoning is unnecessary.