Models by Plan

Tukun.ai documents model access as a plan contract, not as an open-ended provider list.

That distinction matters:

• users buy product capacity, not raw provider access
• the visible model selector only shows models allowed for the current account plan
• the exact provider mix may evolve, while the plan contract should remain understandable

The current contract

Today the runtime model catalog is organized around plan-scoped access plus server-side selection.

The stable rules are:

• free accounts use a smaller entry-level catalog
• pro and max accounts can use Auto
• Auto resolves on the server, inside the active plan only
• the account does not gain unrestricted access to every provider or every model family
• model usage is still priced in unified credits, so stronger models can consume credits faster

Plan matrix

The current product behavior is best understood like this:

Plan	Catalog shape	Default behavior	Auto available
Free	Entry-level plan-scoped models	Uses the plan default model	No
Pro	Paid plan catalog with stronger eligible models	Starts from the Pro default model	Yes
Max	Widest paid catalog and strongest default	Starts from the Max default model	Yes

This table is intentionally written at the contract level. It describes what the user can rely on even if the exact provider-backed entries in the catalog change later.

Current model list by plan

The following list reflects the current runtime catalog actually available in Tukun.ai today.

Plan	Current available models	Default model	Auto available
Free	DeepSeek V4 Flash, GPT-5.6 Luna	DeepSeek V4 Flash	No
Pro	DeepSeek V4 Flash, GPT-5.6 Luna, DeepSeek V4 Pro, GPT-5.6 Terra, GPT-5.5	DeepSeek V4 Pro	Yes
Max	DeepSeek V4 Flash, GPT-5.6 Luna, DeepSeek V4 Pro, GPT-5.6 Terra, GPT-5.5, GPT-5.6 Sol	GPT-5.5	Yes

Additional notes for the current list:

• DeepSeek V4 Flash and GPT-5.6 Luna are the current lighter entry-level models.
• DeepSeek V4 Pro, GPT-5.6 Terra, GPT-5.5, and GPT-5.6 Sol are the current stronger paid-plan models.
• DeepSeek V4 Pro, GPT-5.6 Terra, GPT-5.5, and GPT-5.6 Sol are currently candidates for stronger capability routing.
• Max adds GPT-5.6 Sol beyond the concrete model list available to Pro.
• Max is currently shown in product surfaces as a higher-usage plan preview, so model visibility and sellability should not be conflated.

What the selector means

The model selector is not a marketplace of raw providers.

Instead:

• the selector reflects the account’s current plan entitlement
• the default visible choice may be a named model or Auto
• if a model is not in the plan-scoped catalog, the account should treat it as unavailable

For paid plans, the selector can expose either a concrete model choice or Auto. For free plans, the current product behavior does not allow setting the account model selection to Auto.

How Auto actually resolves

Auto is a server-side routing rule, not a promise that every turn uses the strongest available model.

The current runtime behavior is:

1. Start from the active plan’s default model.
2. If the turn explicitly needs reasoning support and the default model does not support reasoning, move to a reasoning-capable model that is still allowed by the plan.
3. If the turn is an analysis-oriented turn and the plan default is only a lighter entry-level model, upgrade to a stronger allowed model.
4. Otherwise stay on the plan default model.

In short, Auto is conservative by design. It optimizes for predictable everyday usage first, then upgrades only when the request needs it.

What counts as an upgrade trigger

The current documented triggers are:

• the user or runtime explicitly requests reasoning support
• the turn is classified as analysis-grade execution rather than a lighter normal turn

This is important operationally because users may see different concrete models across different turns in the same conversation, even when they did not manually change the selector.

What is stable versus what may change

Treat these as stable product rules:

• plans gate the model catalog
• paid plans can use Auto
• free plans cannot rely on Auto as an account model selection
• server-side routing decides the concrete model for Auto turns

Treat these as implementation details that may change:

• the exact provider names behind a plan
• the exact model names shown in the catalog
• the exact ordering of models inside a paid catalog

If the runtime catalog changes, this page should be updated as part of the same release so the docs remain the public source of truth for current plan-to-model mapping.

How to explain plans to a customer

When a customer asks “which model do I get?”, the best answer is usually:

• Free is for entry-level usage with a smaller model catalog
• Pro adds paid-plan model access plus Auto for reasoning or heavier analysis
• Max starts from the strongest default and keeps the widest current catalog

That explanation is usually more accurate than reciting every provider/model name, because the durable promise is the plan-scoped capability boundary.