Same library. Different brain.
What actually changes between the four tiers?
Every Radiant Vault ships the identical knowledge library: encyclopedia, maps, medical and emergency references, and the full education curriculum. Cheaper tiers don’t get less content. What scales is the AI itself, how large the model is, how fast it answers, and how much nuance it can hold in a single response.
A visual, not a benchmark: bigger chip, denser grid, fuller bar. Same idea underneath as the spec table below.
Two ways to read this
In plain terms, and for the technically curious
In plain terms
Think of the AI like a research assistant who has already read the entire library, on every tier. The difference is how sharp that assistant is and how long you wait for an answer. Cache gives you fast, reliable answers to everyday questions. Vault holds a real conversation and drafts things for you, about as capable as a solid free AI app most people already use. Reserve and Citadel get closer to a genuine research and writing partner, one that almost never makes you wait, and can reason through harder, longer problems.
For the technically curious
Each tier runs a local large language model at a different parameter count, roughly, how much the model learned during training. More parameters generally means better reasoning and nuance, at the cost of needing more compute to stay fast. Response speed is measured in tokens per second: at 5–20 tok/s you’re watching words stream in, at 100+ tok/s a full answer lands close to instantly. Graphics hardware is what makes the difference, from integrated graphics on Cache up to a dedicated RTX 5090 on Citadel.
Full spec comparison
Every number, side by side
| Cache | Vault | Reserve | Citadel | |
|---|---|---|---|---|
| Price | $561.75 pre-order ($749 list) | $1,011.75 pre-order ($1,349 list) | $1,986.75 pre-order ($2,649 list) | $8,136.75 pre-order ($10,849 list) |
| Storage | 500GB SSD | 1TB NVMe | 1–2TB NVMe | 2TB+ NVMe |
| AI model | 1–3B params | 3–8B params | Up to 13B+ | 30B (to 70B) |
| Speed | 5–20 tok/s | 30–55 tok/s | 100–800+ tok/s | 100+ tok/s |
| Graphics | Integrated | Radeon (int.) | RTX 3060+ | RTX 5090 32GB |
Specs are estimates pending a real-hardware validation build.
Practical examples
What this actually feels like to use
Four everyday tasks, and how each tier actually handles them.
| Task | Cache | Vault | Reserve | Citadel |
|---|---|---|---|---|
| Look up a first-aid step from the library | Instant, read straight from the reference. | Instant, plus a follow-up question if you need more detail. | Instant, with nuance and cross-referenced caveats. | Instant and exhaustive, pulled together across related entries. |
| Draft a letter or email | Drafts something short and usable; plan to edit it. | Drafts a full, natural-sounding letter in one pass. | Drafts it, then revises tone or length instantly on request. | Handles nuanced, high-stakes writing with little editing needed. |
| Help with homework | Answers most K-12 questions, kept brief. | Comfortable across the full K-12 curriculum, with real back-and-forth. | Same, at a pace fast enough for a real tutoring session. | Handles advanced coursework too, not just K-12 basics. |
| Summarize a long document you loaded | Short documents only; not really built for this. | Comfortable with a document a few pages long. | Handles a full 40-page manual, answers land as fast as you read. | Reasons across long, dense documents: research papers, full manuals. |
Common questions
Before you pick a tier
Do cheaper tiers get less content?
No. Every tier ships the exact same knowledge library.
- ◆Encyclopedia, maps, medical, and education content are identical across all four
- ◆What's different is the AI model and hardware, not the library
- ◆Content packs are available as add-ons on every tier equally
Can I upgrade after I buy?
Yes. Storage and memory are user-serviceable across all tiers.
- ◆Add storage or memory yourself as your needs grow
- ◆Citadel accepts the Ultimate GPU upgrade for its largest models
- ◆No tier locks you out of future hardware changes
Which tier is right for a household?
Vault is the sweet spot for most households that use it daily.
- ◆Cache suits quick-reference use: lookups, not long conversations
- ◆Vault handles real drafting, research, and homework help
- ◆Reserve and Citadel suit heavy daily use or larger households
Still not sure which one fits?
Reach out and describe how you'd actually use it.
- ◆Tell us your household size and what you'd use it for most
- ◆We'll point you at the tier that actually fits, not the most expensive one
- ◆No pressure, no sales script