How to Use Llama 3 8B Lunaris on Composite
At a glance
| Model ID | sao10k/l3-lunaris-8b |
|---|---|
| Free ID | sao10k/l3-lunaris-8b:free |
| Made by | Sao10K |
| Context window | 8,192 tokens |
| Longest reply | 7,372 tokens |
| Input | Text |
| Reasoning | No |
| Tool calling | No |
Lunaris 8B is a generalist and roleplaying model from Sao10K, based on Llama 3. It is a merge of several models, made to balance creativity with better logic and general knowledge. It is the only model on Composite's free list described in its listing as a roleplay model.
The 8,192-token window decides everything
Lunaris can only see 8,192 tokens at once, counting the character card, the preset, the chat history and the reply. That is roughly 6,000 English words in total. Most setup problems with this model come from sending more than that.
- Set your app's context size to 8,192 or slightly below, so the app trims history before the model has to.
- Set the response length to a few hundred tokens. A long reply takes space from history.
- Keep the character card short. A card of 1,000 tokens leaves room for history; a card of 4,000 leaves very little.
- Prefer a light preset. Large community presets written for million-token models can fill most of the window by themselves.
When the chat grows past the window, Composite trims the oldest messages so the request still fits, and Lunaris stops seeing them. Put facts the character must not forget into a lorebook entry so they come back when a keyword appears.
Samplers
Lunaris accepts the full set of samplers roleplay presets use: temperature, top P, top K, min P, repetition penalty, and frequency and presence penalties, plus stop sequences and logit bias. That makes it a good model for trying sampler changes, since each one has an effect. It does not support tool calling or reasoning.
When to move up
If your story needs a long memory or a large lorebook, the window will keep getting in the way. Mistral Nemo on the same free list has a 131,072-token window, and Synth has 30,000 tokens on its own separate daily count.
Connect it to your chat app
- Create an API key on the API Keys page.
- In SillyTavern, ChubAI, Agnai or RisuAI, choose an OpenAI-compatible provider and set the base URL to
https://composite.lucidity.sh/v1. JanitorAI wants the full URL:https://composite.lucidity.sh/v1/chat/completions. - Paste the key, then type
sao10k/l3-lunaris-8b:freeas the model name. - Send a short message to check the connection before you load a long chat.
Step-by-step setup for each app is in the JanitorAI and SillyTavern guides, and the docs hub covers the others.
From code, it is a normal OpenAI-compatible request:
curl https://composite.lucidity.sh/v1/chat/completions \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "sao10k/l3-lunaris-8b:free", "messages": [{"role": "user", "content": "Hello"}]}'Free access (as of October 21, 2026)
On October 21, 2026, this model is on Composite's free list as sao10k/l3-lunaris-8b:free. The plain sao10k/l3-lunaris-8b stays paid. Every account shares 80 requests a day across all of the free models, separate from Synth's count, and the count resets at midnight UTC. Credits and plans do not raise it.
The free list changes. Check the free models on the Models page for the current set, and the features guide for the rules. A :free id that is no longer on the list returns a model-not-found error and charges nothing. When the 80 requests are used up, requests stop with an error until the reset; Composite never charges your credits to continue. You can switch to sao10k/l3-lunaris-8b on credits or a plan, or to Synth.