Galactre
USD

Choose how catalog prices and estimates appear. Checkout and account records remain in IDR.

Open Chat

GALACTRE MODEL FAMILIES

Find your next
way forward.

Write with clarity. Explore a difficult question. Work through your next piece of code. Get to know the GT family developed by Galactre and choose a model with a clear view of what is available.

Text conversations · Guest and account access · Developer integration

01 / THE FAMILY

A model for the work ahead.

Start with a service you can use today. Explore the direction of the family as it grows. Availability labels describe the current catalog; intended workflows help explain each family.

Planned

GT Lunar

A practical place to start.

Status unavailable
gt-lunar

No released version is available.

Read availability details

Work through questions, drafts and code with a conversational model. Break a large task into focused requests and refine the result with follow-up questions.

Intended workflows

  • Turn notes into a clear first draft, then adjust tone and structure.
  • Ask for explanations, compare approaches and make assumptions explicit.
  • Discuss a code snippet, identify possible edge cases and draft tests.
Context window
Not published
Maximum input
Not published
Maximum output
Not published
Input rate
Not published
Output rate
Not published

Verify facts against your sources and run generated code in your own development environment. A text response does not mean code was executed.

Planned

GT Stellar

More room for connected work.

Status unavailable
gt-stellar

No released version is available.

Read availability details

The balanced family for tasks that connect several pieces of information: structured reviews, longer briefs and explanations that weigh alternatives.

Intended workflows

  • Develop a structured brief from supplied requirements.
  • Compare design alternatives against explicit constraints.
  • Review an argument for gaps, dependencies and trade-offs.
Context window
Not published
Maximum input
Not published
Maximum output
Not published
Input rate
Not published
Output rate
Not published

These are intended workflows. Evaluate the enabled release on representative tasks before selecting it for your application.

Planned

GT Nebular

A direction for deeper analysis.

Status unavailable
gt-nebular

No released version is available.

Read availability details

The frontier family for complex problems that benefit from a structured breakdown, careful assumptions and a detailed review of competing approaches.

Intended workflows

  • Break down a complex problem into verifiable steps.
  • Explore architecture decisions and their operational implications.
  • Build an analysis plan with assumptions, risks and validation criteria.
Context window
Not published
Maximum input
Not published
Maximum output
Not published
Input rate
Not published
Output rate
Not published

This is a product direction. No comparative benchmark, accuracy score or execution capability is promised before release.

Planned

GT Cosmic

Our premium frontier direction. In development.

Status unavailable
gt-cosmic

No released version is available.

Read availability details

GT Cosmic is the premium frontier family in development. Generation, pricing and release specifications are not available yet.

Intended workflows

  • Explore difficult research questions with explicit evidence and validation criteria.
  • Plan complex technical investigations across multiple constraints.
  • Review high-complexity proposals before real-world execution.
Context window
Not published
Maximum input
Not published
Maximum output
Not published
Input rate
Not published
Output rate
Not published

Development status does not establish training provenance or measured superiority. Capabilities and evaluations will be published with a verified release.

VERSION SELECTION

Your model.
Your release.

Choose a version from the dropdown beside each model. Latest follows its promoted default; a numbered version keeps the published product configuration. Recorded answers retain the version that responded.

Use gt-lunar for Latest, or gt-lunar-1 for version 1. An unavailable exact version returns an error instead of switching to another release.

Galactre product IDs are gt-lunar, gt-lunar-1 and future releases such as gt-lunar-1.1. On Avarouter, include the developer namespace: galactre/gt-lunar or galactre/gt-lunar-1. The unversioned ID follows the latest promoted release; it is not a separate model.

Token rates have their own revisions. Evaluation reports apply to the release and conditions actually measured.

02 / AT A GLANCE

Know what you are choosing.

Published specifications from the model catalog. A blank specification is never treated as unlimited capacity or a zero price.

Scroll horizontally to compare all model families.

GT model specifications · follows the version selected for each family
SpecificationGT LunarStatus unavailableGT StellarStatus unavailableGT NebularStatus unavailableGT CosmicStatus unavailable
AvailabilityStatus unavailableStatus unavailableStatus unavailableStatus unavailable
Release stagePlannedPlannedPlannedPlanned
FocusEveryday writing, explanations and focused development tasksProduct direction: balanced reasoning across connected tasksProduct direction: complex analysis and multi-step problem solvingPremium frontier research and complex reasoning workflows
Context windowNot publishedNot publishedNot publishedNot published
Maximum inputNot publishedNot publishedNot publishedNot published
Maximum outputNot publishedNot publishedNot publishedNot published
Input / 1M tokensNot publishedNot publishedNot publishedNot published
Output / 1M tokensNot publishedNot publishedNot publishedNot published

Availability is separate from permission and service health. No comparative speed, quality or benchmark ranking is implied by a family name.

03 / MAKE A GOOD START

Choose by the task.
Validate with your work.

For a first conversation, choose an available model and keep the task focused. For an application, test with representative inputs and define what a useful result looks like.

About model evaluations
  1. 01

    Check the release.

    Confirm availability, release stage and published limits. A roadmap family cannot be unlocked by changing a plan.

  2. 02

    Give the right context.

    State your goal, audience, constraints and preferred output. Include the relevant facts rather than assuming access to your files or repository.

  3. 03

    Inspect the result.

    Check citations against your sources, review assumptions and run suggested tests yourself. Measure quality against your own acceptance criteria.

  4. 04

    Refine and keep control.

    Use follow-up messages to improve a draft. Keep an eye on quota, save important results and review access before integrating into a workflow.

04 / FROM AN IDEA TO A PROMPT

Bring something to the conversation.

Explore a starting prompt, copy it, or open it in chat. These are editable examples; no request is sent until you choose to submit it.

A clearer draft starts with a better brief.

Provide your audience, goal and tone. Ask for a first draft, then refine specific passages in the same conversation.

Check that the retry count and behavior match your implementation before publishing.

Use this prompt in chat
Example prompt · editable in chat
Rewrite the following release note for developers. Use a clear, direct tone and keep it under 120 words. Preserve the technical details. Flag any information that is missing instead of inventing it.

Draft: We added retry support to the background job runner. Failed jobs can retry up to three times. The admin can inspect each attempt in the activity log.

05 / FOR DEVELOPERS

A clear identity.
A documented contract.

Use gt-lunar, gt-stellar and gt-nebular as the product identifiers when inspecting the catalog. Browser requests accept model with a family ID or an exact release ID. The older family: "Lunar" shorthand follows the current default.

The example uses a same-origin browser session. It keeps credentials out of your JavaScript and sends the CSRF token returned by the session endpoint.

JavaScript · same-origin browser
const catalog = await fetch('/api/v1/chat/models')
  .then(response => response.json());
const lunar = catalog.models.find(model => model.family === 'Lunar');
if (!lunar?.available) throw new Error('Lunar is not currently available.');

const session = await fetch('/api/v1/chat/session', {
  credentials: 'same-origin'
}).then(response => response.json());
if (!session.chat_available || !session.csrf_token) {
  throw new Error('Chat is temporarily unavailable.');
}

const response = await fetch('/api/v1/chat/generate', {
  method: 'POST', credentials: 'same-origin',
  headers: {
    'Content-Type': 'application/json',
    'X-CSRF-Token': session.csrf_token
  },
  body: JSON.stringify({
    family: 'Lunar', model: 'gt-lunar', message: 'Help me review this approach.'
  })
});
const result = await response.json();
if (!response.ok || result.status !== 'completed') {
  throw new Error(result.error?.message || 'Request unavailable.');
}
console.log(result.public_identity, result.response);

06 / UNDERSTAND THE BOUNDARIES

Your session. Your context. Your choice.

Model specifications and application limits answer different questions. Review both before starting a conversation or making a purchase.

Try as a guest

Guest access is enabled for Lunar, subject to current service availability.

Requests per guest / day
20
Characters per message
2,000
Session lifetime, up to
24 hours

Daily limits reset in UTC. Additional rate and network limits apply. Remaining quota is shown in chat.

Try a conversation

Use your account

Sign in for separate account history, credit purchases and authorized workspace features. Your model access remains subject to the catalog and your account permissions.

  • Guest and account history remain separate.
  • Review credit packages and invoices before paying.
  • A purchase does not enable an unreleased model.
Log in to your workspace

Understand conversation context

Recent messages help you continue the conversation. The application includes a bounded selection of up to 510 recent messages, with additional size limits.

This is separate from a model context window. A long chat is not a guarantee that every earlier message is used. Restate essential requirements when the topic changes.

Read about context

Work with care

Conversation text is encrypted at rest and access is scoped to the owning guest or account. Delete saved conversations through the history controls.

Do not include passwords, API keys or material you are not authorized to share. Treat generated answers as drafts to review, especially for decisions with significant consequences.

Privacy and data handling

07 / A FEW MORE ANSWERS

Before you start.

Capabilities, availability and billing, explained in one place.

How do family IDs and version IDs differ?

A family ID such as gt-lunar follows the current administrator-approved default. An exact ID such as gt-lunar-1 pins that published product release. Creating a draft does not move the default. Unavailable or retired versions return an error instead of silently switching.

Which model can I use now?

Use the availability labels on this page. Only models marked Available can be considered for execution; your session, limits and current service health still apply. Preview identifies the release stage and does not promise an uptime or accuracy guarantee.

Can I try Galactre without an account?

When guest access is enabled, you can start a Lunar conversation without signing in. Guest requests are limited and conversations belong to that browser session. Sign in for separate account history, credit purchases and workspace access. Guest history is not automatically transferred into your account.

Are Stellar and Nebular available through another plan?

A plan or credit purchase does not unlock an unavailable model. These families become selectable only after their service is enabled and your account has permission. Intended use cases describe product direction, rather than measured performance.

What do context and output limits mean?

The context window is the token capacity published for a model. The output limit is the maximum generated answer length. A token is not the same as a character. Chat also applies its own input and conversation-context limits, so a model limit is not a promise that an entire conversation will be included. Unpublished limits are shown explicitly.

Does Galactre browse, run code or generate images?

Chat accepts and returns text. Signed-in conversations can connect private knowledge sources and enable the bounded calculator and current-time tools. File uploads, image generation, browsing and shell execution are not released. Verify important answers against their cited sources.

How do model prices relate to credits?

Published input and output rates, when provided by the catalog, are shown per million tokens. They are separate from a credit package purchase. Review the amount, currency and credit quantity in your authenticated billing workspace before checkout. An unpublished rate does not mean free or unlimited usage.

Where can I find evaluations or release changes?

The evaluation page explains the current publication status. Read the actual published evaluation reports and their measurement scope. An internal development result is not a general superiority claim. Review catalog availability and limits again before a new deployment; preview behavior and availability may change.

YOUR NEXT QUESTION STARTS HERE

Put an idea into words.

Start with an available model, bring your context, and build from the answer.

Open GT Chat