Free models
Two free models, always. The credit system never gates the basic right to chat.
Reverie's commitment: there will always be at least one fully free LLM available to every user. No daily quota, no rate limit beyond shared-pool fairness, no upgrade prompt.
Source: The free models commitment →
The free models today
| Model | Type | Context | Notes |
|---|---|---|---|
| Llama 3.1 8B | LLM, text-only | 131K | General conversation |
| Qwen3 VL 30B | LLM with vision | 32K | Image understanding |
Both are 0× multiplier — every message you send on them costs zero credits.
What free means in practice
- Zero per-message cost. Your credit balance doesn't move when you chat on a free model.
- All features work — forking, memory, personas, plugins, group chats, voice (during TTS promo). The free model isn't a stripped-down product, just a cheaper one.
- NSFW supported — content filter rules apply equally.
- Same uptime as paid models.
The free-tier rate limit
Free LLMs run under a dynamic rate limit per user — designed to keep them genuinely useful for everyday chat while preventing automation and quota-stretching.
- It's a sliding window, not a daily bucket. Slots refill gradually as old messages age out, not all at once at midnight.
- Shared across surfaces. Web chat, story mode, novel mode, Discord/Telegram bots, and the Open API all share one quota. Switching surfaces doesn't reset it.
- Tunable. We adjust the limit over time based on usage patterns and abuse signals — we don't publish a fixed number because it'll change.
When you hit the limit, the chat tells you how long until the next slot opens up. You can either wait, or switch to a paid model (any non-zero multiplier in the model picker) and keep going.
What's the catch
The free models are smaller than the paid models. Expect:
- Shallower replies, especially in emotionally complex scenes
- More repetition in long chats
- Slightly weaker character voice fidelity (they tend to drift)
- Reduced complexity of inference (they sometimes miss subtext)
For everyday chat — fine. For your most-loved characters in their most important arcs — switch to a paid model for those scenes.
Why we commit to free
Three reasons:
- Accessibility. Many users are in regions where any paid subscription is expensive relative to local incomes. Free models keep Reverie usable globally.
- Trust. A platform that uses every interaction to monetize you ends up monetizing your sentimentality. Free models are a hard-line "no" against that pressure.
- Discovery. Most users start free and decide what they care about before they decide what to pay for. Free models are the on-ramp to paid features.
Will the specific free model change
Yes. We rotate the free model when a better small/cheap model becomes available. The commitment is to a free model, not to Llama 3.1 8B specifically — when the catalog changes, the model picker is the source of truth for what's currently free.
Mixing free and paid
You can absolutely use free for casual chat and switch to paid for important scenes — that's the recommended pattern. The model picker is instant. Memory and conversation history are model-agnostic; switching doesn't restart anything.
Quality comparison
The chasm between free and paid is real. From a typical romance scene:
Llama 3.1 8B reply:
GLM 5 reply:
For most paid features, the free model is enough. For good writing, it isn't.