Model Selector

The Yukkai model selector lets you choose from over 300 LLMs — including Yukkai's own Fusion, Frontier, Free, and Local options — with a personalized favorites list.

Yukkai can work with many AI models. Models differ in reasoning ability, speed, price, context size, vision support, and tool-use reliability. The model selector is your control center for choosing the right model for every situation.

How the Model Selector Works

The model picker is located in the chat interface. Click it to see your personalized model list. From there you can:

  • Select a model — switch the active model for the current conversation.
  • Reorder favorites — drag models to arrange them by preference.
  • Remove models — hide models you don't use from your list.
  • Add models — browse the full catalog and add models to your list.
  • Search — find any of the 300+ available LLMs by name or provider.

Model Categories

The catalog is organized into categories to help you pick the right model:

Yukkai Meta-Models

These are Yukkai's own routing identities — you pick the tier, and Yukkai automatically routes each message to the best underlying model:

Model Best For Cost Billing
Yukkai Free Light everyday requests, trying Yukkai Free (daily/monthly limits) No charge
Yukkai Fusion Everyday use, balanced quality/cost Lower Yukkai Wallet
Yukkai Frontier Complex reasoning, advanced coding, deep research Higher Yukkai Wallet
Yukkai Local Offline use, privacy, no cost Free No charge

Note: Fusion, Frontier, and Free require a Yukkai account key (yk_...). They are not available when using a personal OpenRouter key.

Personal Cloud Models

With a personal OpenRouter key, you can select individual models from multiple providers (Anthropic, OpenAI, Google, DeepSeek, Meta, Mistral, and more). Choose a personal model when you want direct control over:

  • Provider — pick a specific vendor.
  • Model identity — know exactly which model answers.
  • Price — see per-token costs upfront.
  • Context size — pick models with larger context windows.
  • Speed — choose faster, lighter models for simple tasks.
  • Specialized capabilities — vision, code, reasoning, etc.

Availability and pricing are controlled by OpenRouter and the underlying provider. The catalog includes over 300 LLMs and is updated daily as new models are released and others become unavailable.

Local Models (MLX)

Run open-weight models entirely on your Mac via Apple's MLX framework. The picker features a curated shortlist of ~8 recommended models (Llama 3.2 3B, Qwen3 4B, Qwen3-Coder-30B, Gemma 4 27B, etc.), each tagged with:

  • Parameter count
  • RAM requirement
  • Download size

You can also enter any Hugging Face model id (e.g. mlx-community/DeepSeek-R1-Distill-Qwen-7B-4bit) to download hundreds more MLX-converted models. See Yukkai Local (MLX) for details.

Important: Local models have significant limitations — no tools, no vision, no streaming. They're plain text completion only.

Choosing the Right Model

Here's a quick decision guide:

Task Recommended Model
Quick question, light task Yukkai Free or Yukkai Fusion
Everyday coding, writing, research Yukkai Fusion
Complex reasoning, architecture, deep analysis Yukkai Frontier
Privacy-sensitive, offline work Yukkai Local
Specific provider/model preference Personal OpenRouter model
Image/music/video generation Always uses Yukkai's paid Google models (regardless of chat model)

Model Capabilities

Each model in the catalog shows its capabilities:

  • Vision — can process images.
  • Tools — supports tool calling (required for most Yukkai features).
  • Context window — maximum tokens the model can handle.
  • Speed — relative inference speed.

Tip: For Yukkai's tools to work (file editing, web browsing, desktop automation, swarm, etc.), the selected model must support tool calling. Local MLX models do not support tools.

Switching Models

You can switch models at any time — even mid-conversation. The new model takes effect on your next message. This is useful for:

  • Starting with Fusion for exploration, then switching to Frontier for deep analysis.
  • Using Free for simple questions and Fusion/Frontier for complex ones.
  • Testing how different models handle the same prompt.

Custom Model Lists

You can maintain a personalized list of models:

  1. Open the model picker.
  2. Click Edit List or Manage Models.
  3. Add models from the catalog, reorder by dragging, or remove models you don't use.
  4. Your list is saved and synced across sessions.

Adding API Keys for Direct Model Access

Beyond the model picker, you can add provider API keys (Replicate, Fal.ai, etc.) to the Vault and have Yukkai call a specific model directly for a task. For example, register your Replicate key to access dozens of video generation models.

Good to Know

  • Media generation is separate — image, music, and video generation always use Yukkai's paid Google models, regardless of which chat model is active.
  • Model availability changes — the catalog is updated daily as providers add or deprecate models.
  • Not all models are tested — Yukkai curates and recommends models that work well, but the full 300+ catalog includes models we haven't verified.
  • Costs vary — check the Wallet or OpenRouter pricing before committing to an expensive model.

Learn More

Last updated: July 19, 2026