Supported models
Reverie's current LLM, image, and video models, organized by tier. Multipliers and the lineup itself move over time — this page is a snapshot.
Reverie supports a portfolio of LLM, image, and video models. We add new ones as providers ship them and retire older ones when their replacements are clearly better. The in-app model picker is always the source of truth.
The lineup and multipliers change over time. Models get added, retired, or repriced as the underlying provider ecosystem moves. A model that's "advanced" today may be a basic option next quarter, and entirely new models land regularly. The in-app model picker is the authoritative source for what's available right now.
LLM models
Two tiers:
- Basic — available on every account, free and paid.
- Advanced — gated to active subscribers. The model picker shows a lock and the chat refuses to switch if you don't have a plan.
Basic tier
| Model | Multiplier | Context | Reasoning | Notes |
|---|---|---|---|---|
| MiMo V2 Flash | 0.3× | 262K | Optional | Default for new users. Fast and cheap. |
| DeepSeek V4 Flash | 0.3× | 163K | Optional | Cheap and conversational. |
| DeepSeek V3.2 | 0.5× | 163K | — | Reliable workhorse for roleplay. |
| DeepSeek R1 | 1.0× | 163K | Always | Pure reasoning model. |
| GLM 4.7 | 1.0× | 200K | Optional | Balanced quality. |
| Llama 3.1 8B | 0× — free¹ | 131K | — | Always-free text model. |
| Qwen3 VL 30B | 0× — free¹ | 32K | — | Free model with vision. |
¹ Free models share a dynamic per-user rate limit across all surfaces (web, bots, API). See Free models.
Advanced tier (subscribers only)
| Model | Multiplier | Context | Reasoning | Notes |
|---|---|---|---|---|
| Gemini 3 Flash Preview | 1.0× | 1M | — | Million-token context. Multimodal. |
| GLM 5 | 1.5× | 200K | Optional | Smarter sibling of GLM 4.7. |
| GLM 5.1 | 2.0× | 200K | Optional | Highest-quality GLM. |
Reading the columns:
- Multiplier — credits per token. Multiply by reply length to estimate cost.
- Context — the model's max input window. Longer chats need bigger context.
- Reasoning — "always" = always shows thinking; "optional" = you toggle; "—" = no support.
Image models
| Model | Cost / generation | Type |
|---|---|---|
| Z Image Turbo (default) | 5,000 credits | Text-to-image |
| Z Image Turbo I2I | 8,000 credits | Image-to-image (variation) |
Image generation is fast (under 30 seconds typically) and produces illustration-style art tuned for Reverie's image guidelines.
Video models
| Model | Cost / generation | Type |
|---|---|---|
| SeeDance I2V | 120,000 credits | Image-to-video |
Subscriber-only. Short clips animated from a still character image. See Image-to-video.
Model selection in the UI
The model picker sits at the top of every chat. Models you've used recently show first; the full list expands with the chevron. Each entry shows: name, multiplier, context length, and a short description. Tap to switch — the new model takes effect on the next reply.
Advanced-tier models show a lock badge if your account doesn't have an active subscription.
Reasoning models
Three flavours:
- Always-reasoning — the model always emits a visible thinking pane (DeepSeek R1 currently).
- Optional — toggle reasoning on or off per chat (MiMo V2 Flash, DeepSeek V4 Flash, GLM 4.7/5/5.1 currently).
- No reasoning — plain replies only.
Defaults
- New users' default LLM: MiMo V2 Flash.
- New users' default image model: Z Image Turbo.
- Per-character override: creators can set a recommended model on their character; you can still pick any model the chat allows.
Model retirement
When a model retires, conversations that were using it migrate to a successor model automatically — context carries over, you don't lose anything. We don't delete past conversations because a model went away.
What model should you pick?
See Choosing a model for the decision guide.