Skip to content
Models

Every frontier lab,
one composer.

Eleven chat models from eight labs, chosen per conversation — plus 10 image and 10 video models in Studio. Claude Opus 5.5 is the default.

Get early accessSee pricing
AnthropicOpenAIGooglexAIMoonshot AIAlibabaZ.aiDeepSeek
Chat models

Eight labs,
eleven models.

Available in chat, Build, Research and Games — in the order the composer lists them.

Anthropic

3 models
  • Claude Opus 5.5Defaultanthropic/claude-opus-5.5
  • Claude Opus 5anthropic/claude-opus-5
  • Claude Fable 5.1anthropic/claude-fable-5.1
ReasoningLow · Medium · High · Off

OpenAI

2 models
  • GPT-6 Astraopenai/gpt-6-astra
  • GPT-5.6 Solopenai/gpt-5.6-sol
ReasoningLow · Medium · High

Google

1 model
  • Gemini 3.8 Flashgoogle/gemini-3.8-flash
ReasoningThinking · Off

xAI

1 model
  • Grok 4.6x-ai/grok-4.6
ReasoningLow · High

Moonshot AI

1 model
  • Kimi K3moonshotai/kimi-k3
ReasoningThinking · Off

Alibaba · Qwen

1 model
  • Qwen3.8 Maxqwen/qwen3.8-max-0902
ReasoningThinking · Off

Z.ai

1 model
  • GLM 5.3z-ai/glm-5.3
ReasoningThinking · Off

DeepSeek

1 model
  • DeepSeek V4.1 Flashdeepseek/deepseek-v4.1-flash
ReasoningThinking · Off
How it works

Three things happen
when you pick a model.

  1. 01

    Pick it in the composer

    The model menu sits next to the send button in chat, Build, Research and Games. Your choice stays with the conversation.

  2. 02

    Set how hard it thinks

    Each maker exposes its own reasoning choices — Claude Low, Medium, High or Off; OpenAI three efforts; Grok low or high; the others think or don't.

  3. 03

    Checked with the provider

    Availability and reasoning options are asked of the provider each time, never assumed. If a model is down, you're told and can switch.

Reasoning,
per maker

MakerChoicesBehaviour
AnthropicLow · Medium · High · OffExtended thinking is optional, at three budgets.
OpenAILow · Medium · HighReasoning models always reason, at three efforts.
GoogleThinking · OffThinks or doesn't.
xAILow · HighTakes low or high effort only.
Moonshot AIThinking · OffThinks or doesn't.
Alibaba · QwenThinking · OffThinks or doesn't.
Z.aiThinking · OffThinks or doesn't.
DeepSeekThinking · OffThinks or doesn't.

These are each maker's usual choices, used when the provider's catalog can't be read. When it can, the model's own list is used instead.

Studio has
its own lineup

10image models
10video models

Image and video models are chosen in the Studio panel, which only offers the options each model accepts — aspect ratios, resolutions, clip lengths and sound.

Explore Studio

What we
measured

The published benchmark ran on Claude Opus 5.5 only, on one machine, so that the engine was the only variable. Other models were not measured there.

Read the benchmarks

Questions

Which model is the default?
Claude Opus 5.5. You can change the model for any conversation from the composer.
Can I switch models mid-project?
Yes. The choice belongs to the conversation, so a new conversation — or a new run — can use a different model.
Why does a model's reasoning menu differ from this page?
The app asks the provider's catalog for each model's own reasoning list and uses it when it can. The choices on this page are the fallback used when that list can't be read.
Were all models benchmarked?
No. The published benchmark used Claude Opus 5.5 only, on one machine. Other models were not measured there.
Are Studio models chat models too?
No. Studio's image and video models are a separate lineup, chosen in the Studio panel.

Every frontier model.
One app, one thread.

Get early accessSee pricing