TSK-1 Model Tiers

TSK-1 (Taskade System Kernel) is the intelligence layer that auto-routes AI work across four capability tiers — vendor-agnostic, always current.

TSK-1 is not a single model. It is an orchestration layer — the intelligence system that coordinates how AI models, memory, agents, and workflows operate inside your workspace. When you pick a tier, TSK-1 routes your request to the best available model for that capability level.

The short version. Set your agent to Auto and TSK-1 picks the right tier per request. The four tiers — Instant, Standard, Thinking, Reasoning — describe what the system is good at, not which provider powers it.

---

How TSK-1 fits in the stack

Taskade → Genesis → EVE → TSK-1 → Workspace DNA → Running workspace

TSK-1 sits between the EVE persona and Workspace DNA — it is the engine that decides how to execute a given request across the Memory (projects), Intelligence (agents), Execution (automations), and Apps (Genesis) pillars.

---

The four tiers

The model picker shows one TSK-1 header with a capability ladder. Each tier is defined by what it's good for, not by its underlying provider:

Tier	Best for	Example use
Instant	Fast edits, simple completions, quick lookups	Rename a task, fix a typo, categorize an item
Standard	Everyday work — balanced speed and reasoning	Draft an email, summarize a document, answer a question
Thinking	Planning, creation, multi-step reasoning	Design an automation flow, write a project proposal, generate a Genesis app
Reasoning	Complex systems, deep analysis, multi-agent coordination	Audit a workspace, cross-reference databases, run a multi-agent team

---

Auto (recommended)

Set your agent to Auto and TSK-1 dynamically routes each request to the appropriate tier — fast when you need speed, deep when you need reasoning. Auto is the default for new agents on most paid plans.

Agents or automations pinned to a specific tier stay on that tier until you change them. The tier definitions are stable; the underlying models that serve each tier may change as the frontier improves.

Model lineup may change. TSK-1 runs on the latest frontier models. Because the tiers describe capability rather than provider, Taskade can upgrade the underlying models without you needing to reconfigure anything. The in-app model picker always shows the current tier-to-model mapping.

---

When to pin a tier vs. use Auto

Scenario	Recommendation
General-purpose agent — chat, answer questions, draft content	Auto — let TSK-1 decide
High-volume automation that needs low latency and low cost	Instant or Standard
Agent that designs workflows, apps, or complex plans	Thinking
Agent that audits, cross-references, or coordinates other agents	Reasoning
You have a specific model preference from prior testing	Custom model (available on Max and Enterprise plans)

Agents pinned to a custom model (e.g. a specific Claude or GPT version) bypass TSK-1 routing entirely — they always use that model. This is available on Max and Enterprise plans.

---

Credits

Each tier has its own per-token credit cost, shown in the in-app model picker. Auto routes to the tier it selects, billing at that tier's rate. Pinned custom models bill at the model's own rate.

See Credits & Billing for current rates.

---

Workspace DNA

TSK-1 powers all four pillars of Workspace DNA:

1. Projects remember — your data and context stay connected.

2. Agents reason — multi-agent teams understand goals and take action.

3. Workflows run — automations continue after chat ends.

4. Apps launch — one prompt, one live app.

For the full architecture, see How Taskade Genesis Works.

---

What's next

Guide	What you'll learn
AI Agents Getting Started	Set up your first agent and choose a TSK-1 tier
Agent Knowledge & Memory	Give agents context from your projects
Multi-Agent Collaboration	Coordinate multiple agents with TSK-1 Reasoning
Credits & Billing	Understand credit costs per tier