Polaris LLM Guide
Polaris is Kindroid's newest language model. In the app it's described as having stronger memory, richer character voice, and more attentive, immersive conversations. This guide covers who can use it, how to turn it on, and how its settings work.
Who can use Polaris#
- Paid subscribers: Polaris needs an active paid subscription. It works on Standard, Ultra and MAX.
- 3-day trial: You can select Polaris during the trial, but replies come from Ember until the trial converts to a paid subscription.
- Free accounts: The LLM Model list only shows Latest (Ember), and free preview messages always use Ember.
For plan details, see Subscriptions.
Turning on Polaris#
Location: In a chat, tap the ••• menu in the upper right and choose Change Model. You can also open your Kindroid's settings (tap their name at the top of the chat, or the gear icon on desktop) and go to Models.
- Open Models for the Kin you want to switch.
- Tap LLM Model and choose Polaris.
The change saves right away, and you'll see a Saved message.
Note: Polaris is opt-in. New Kins start on Latest (Ember), which runs Ember, not Polaris. A Kin you add from a shared Kindroid keeps the model its creator picked, so check Models if you're not sure which one it uses.
The model is set per Kin, so you can try Polaris on one Kin and keep the rest on Ember. A Kin set to Polaris uses it in group chats too. In a group chat, tap that Kin's avatar and choose Change Model to reach the same settings.
Polaris settings#
When Polaris is selected, the Models page shows LLM Flair, Chat Dynamism and Reasoning Effort. If you switch an existing Kin to Polaris, its dynamism and reasoning settings stay the same.
LLM Flair#
Polaris has four flairs. Companion is the default.
- Companion: Warm and conversational. Less writing, more natural talking.
- Roleplay: Character-focused. Stays in character and brings out their traits.
- Minimal: Bare-bones system instructions, so you get the most control over your own setup.
- Narrative: Storytelling mode, where the AI drives the story with descriptive, vivid prose.
Reverie's Classic flair isn't available on Polaris. If you switch a Kin that uses a flair Polaris doesn't have, it moves to Companion.
Chat Dynamism#
The Chat Dynamism slider goes from 0.60 to 1.80, and the default is 0.95. Lower values keep replies more stable and predictable. Higher values make replies more creative, but also more chaotic. Tap Set to default to go back to 0.95.
Reasoning Effort#
This slider controls how much Polaris plans before it replies. The default is Speedy.
- Speedy: Quick replies with very little hidden planning. Best for spontaneous, natural chat.
- Moderate: A deeper reasoning pass for more reliable answers, while still replying quickly.
- Slow: More reasoning to keep track of context, but replies take longer.
- Very slow: Deliberate reasoning for more depth and fewer mistakes.
- Extra slow: Extended reasoning for the most complex scenarios. It's a lot slower, so save it for when you really need it.
If the planning step fails, Polaris still sends a normal reply without it.
Context and memory#
How much Polaris can keep in view depends on your plan. Standard gives it the base amount, Ultra gives it more, and MAX gives it the most. That space holds your recent conversation and Cascaded memory. You can compare plans under Profile > Billing, and learn how the different kinds of memory work in Memory.
Limits to know about#
- Reply length: Polaris replies max out at 400 tokens, which is shorter than Ember's 600. If a reply stops partway through, tap your Kindroid's latest response and choose Continue.
- Service status: If Polaris seems slow or isn't replying, check the Kindroid Status page. Polaris is listed under Text (LLM).
Tips#
- Start with the defaults: Companion, Speedy and a dynamism of 0.95 are a good baseline. Only raise Reasoning Effort if you need more depth, since each step up makes you wait longer for replies.
- Make small dynamism changes: If replies feel flat, raise dynamism a little. If they get chaotic, lower it or tap Set to default.
- Change one thing at a time: Give each change a few turns to settle before you adjust something else, so you can tell what's working.
For setup advice that works on any model, see General LLM Tips.