Taskade Genesis routes every prompt, agent, and automation through 15+ frontier models from OpenAI, Anthropic, and open-weight providers. Pick a model per task, or let Auto mode decide. No vendor lock-in. Nothing to set up, and no rebuilds when a new model ships. To run on your own provider account, bring your own key instead.
The two-line summary. Premium frontier models for the hardest 20% of work. Open-source for the routine 80%. Auto mode handles the routing. Want the deep dive on the open-source side? See 9 Best Open-Source AI LLMs in 2026, Ranked for Real Work.
Auto mode is TSK-1. TSK-1, the Taskade System Kernel, is the intelligence layer that coordinates 15+ frontier models from OpenAI, Anthropic, and open-weight providers into one running app. Most of the time you never have to pick a model. TSK-1 picks for you, per step, though you can still choose one yourself whenever you want. It is a kernel, not a from-scratch foundation model. See Introducing Taskade TSK-1 for how it works.
TL;DR: Taskade Genesis mixes 15+ frontier models from OpenAI, Anthropic, and open-weight providers in one app. Use lightweight models for cheap iterations and reasoning models for hard logic. Use long-context models for big documents. Taskade adds new frontier models as they ship, so your apps stay current with no rebuild. Build an app →
▲ Memory feeds the prompt
■ Intelligence picks the model (Auto or manual)
● Execution runs it through the gateway
Available Models
Taskade routes every step through 15+ frontier models from OpenAI, Anthropic, and open-weight providers. Pick by job, or let Auto mode choose.
| Capability tier | Best for |
|---|---|
| Fast and cheap | Quick iterations, simple text, high-volume steps |
| Balanced everyday | Code generation, structured output, general reasoning |
| Deep reasoning | Hard logic, math, multi-step planning |
| Long-context | Big documents, long-context analysis, nuanced tasks |
| Open-weight value | Routine bulk routing, cost-sensitive automations |
How to Choose a Model
For Taskade Genesis Apps
Taskade Genesis automatically selects the best model for app generation. You can also specify a model preference in your prompt:
- Complex apps with heavy logic → Use reasoning-focused models
- Design-heavy apps → Use multimodal models that understand visual context
- Quick prototypes → Use fast generation models for rapid iteration
For AI Agents
Each AI agent can use a different model:
- Open your agent settings
- Select the AI model from the dropdown
- Save. The agent now uses your chosen model
💡 Tip: Create specialized agent teams, and give each agent the model that fits its role. A research agent can use one model while a writing agent uses another.
For Automations
The Ask AI and Generate with AI steps show the same full model menu as the rest of Taskade. An automation can reach any model your agents can. Leave the picker on Auto, and Taskade routes the step for you. That is the right answer most of the time.
Three steps keep a shorter, curated list on purpose: Categorize with AI, Respond with AI, and Prompt AI. Those do small, predictable jobs where a hand-picked option beats scrolling the whole catalog.
You can mix models inside one automation. Route a cheap step with a light model, then hand the final answer to a reasoning model.
Why Multi-Model Matters
✓ No vendor lock-in. Switch models without switching platforms.
✓ Best tool for each job. Match the model to the task.
✓ Future-proof. New frontier models are added automatically as they ship.
✓ Clear costs. The model picker shows credit cost per option before you run.
✓ Cost optimization. Use lighter models for simple tasks, and stronger models for complex ones.
Want a deep dive on the open-source picks? See 9 Best Open-Source AI LLMs in 2026.
Credits
AI operations consume credits based on the model used. Check your usage:
- Go to Settings → AI Usage
- View credit consumption by model and operation type
- Upgrade your plan for more credits if needed
See AI Models & Credit Costs for detailed pricing.
Bring Your Own Provider Key
You never need your own AI account. If you already have one, Taskade can run on it instead, and work served that way does not draw Taskade AI credits. There are two separate ways to do this, and picking the right one saves a lot of confusion.
| Workspace default key | Provider steps in automations | |
|---|---|---|
| What it is | One key saved on the workspace that quietly serves your normal agents and AI steps | Dedicated automation steps for a single provider, connected to your own account |
| Where you set it up | Workspace settings, once | On the step, like any other connection |
| What it changes | Nothing visible. The same agents and steps just run on your key | You choose these steps deliberately instead of the built-in AI steps |
| Providers | Google Vertex AI, OpenAI, Anthropic | OpenAI, Anthropic, Google Gemini, Groq, OpenRouter |
| Availability | Enterprise, plus AppSumo and PitchGround lifetime plans | Every plan, including Free, which runs up to 10 automations a month |
| Full walkthrough | Below | Bring Your Own AI Key |
Set a workspace default key
Save a key for a provider in your workspace and eligible agent chats and AI automation steps start using it on their own. Nothing to pick per step.
It applies when all three are true.
- You chose a specific model rather than leaving the step on Auto. Auto stays on Taskade's own capacity, because Taskade picked the model, not you.
- Your plan includes workspace keys.
- Your workspace holds an active key for that model's provider. Google models are served through Google Vertex AI, so a Google service account covers them.
If a saved key cannot be used, for example it was revoked at the provider, the run does not fail. It quietly falls back to Taskade's own capacity and finishes, with credits applying as normal. A past-due account is still blocked either way, since your own key covers the model bill, not your Taskade subscription.
Use a provider step in an automation
When you want one specific step on your own account and the rest on Taskade, add a provider step instead. Connect your OpenAI, Anthropic, Google Gemini, Groq, or OpenRouter account on the step, and that step runs on your key while everything around it keeps working normally. These are ordinary automation steps, so they sit alongside your other actions and you can see exactly which ones use your account.
This is the better choice when you want a single high-volume step off your Taskade credits, or when you need a provider Taskade does not route to natively. See Bring Your Own AI Key for the step-by-step setup.
Choosing the Right Model
Taskade Genesis lets you route any agent or app to 15+ frontier models from OpenAI, Anthropic, Google, and open-weight providers. Or let Auto mode pick for you. Different jobs benefit from different models.
| Capability tier | Best for |
|---|---|
| Balanced everyday | General reasoning, code generation, tool calling |
| Deep reasoning | Hard reasoning, math, multi-step planning |
| Long-context | Long-context analysis, writing, long documents |
| Fast multimodal | Text + image understanding, cheap high-volume tasks |
| Open-weight value | Routine steps, bulk routing, cost-sensitive automations |
Credit costs and plan limits are documented in Model Credits. Auto mode selects the model for you. You can also mix providers inside a single automation. Use a cheap open-source model for a routing step and deep reasoning for the final answer.
Related guides
- Model Credits. Per-model credit costs and monthly quotas.
- Custom AI Agents. Per-agent model selection.
- AI Action. Run a model as a step inside an automation.
- Bring Your Own AI Key. Connect your own provider account to an automation step.
- Taskade EVE. The built-in assistant that routes through all 15+ models.
- MCP Connectors. Let external AI clients like Claude Desktop or Cursor reach into your workspace over MCP.