omne --private-betaheads up → Omne's private beta opens soon[ request a demo → ]

Ask AI

Which AI providers and models Ask AI can use

The text providers Ask AI can run on, how Omne picks the starting model, and why some providers connect but do not appear in the chat.

Ask AI can run on fifteen text providers: Anthropic, OpenAI, OpenRouter, Google Gemini, xAI, Mistral, DeepSeek, Perplexity, Cohere, Together AI, Fireworks AI, Cerebras, Moonshot (Kimi), Qwen (Alibaba Cloud), and any provider that speaks the OpenAI API. Each runs on your own account there.

How Omne talks to each provider

Omne uses two shapes. Anthropic is reached through its own Messages API. Every other provider is reached through the OpenAI-style chat interface that most providers offer, with Gemini going through Google's OpenAI-compatible address. Both shapes stream the answer as it is written and both let the model call tools, which is how a chat reaches your connected accounts.

The models you see

When you connect a provider, Omne asks it for the list of models your key can use. From that list it hides the ones that do not chat: image, voice, embedding, moderation and reranking models, among others. What remains appears in the model menu under the provider's name, with a green Connected marker beside it. Models are sorted by name, with the starting model first.

The model a chat starts on

Omne picks the starting model for you, preferring the provider's strongest everyday model and the newest version of it.

ProviderStarts on
AnthropicThe newest Claude Sonnet, then Opus
OpenAIThe newest GPT-5 model, then GPT-4.1, then GPT-4o
OpenRouterThe newest Claude Sonnet it lists, then GPT-5 models
Google GeminiThe newest Gemini Pro, then Flash
xAIThe newest Grok
MistralMistral Large, then Medium
DeepSeekDeepSeek Chat
PerplexitySonar Pro
MoonshotKimi K2
QwenQwen Max, then Plus

For the other providers, Omne starts on the first chat model the provider lists. You can change the model at any time from the model menu, and Omne remembers your choice on that device.

Perplexity answers without tools

Perplexity answers from the live web by itself and does not take tools. A chat on Perplexity can cite the web, but it cannot call your connected accounts, even when connectors are switched on.

Providers that connect but do not appear in Ask AI

Some providers can be connected in AI providers and still do not show up in the chat's model menu, for one of these reasons.

  • Image, video and voice providers such as ElevenLabs, Runway, Luma, fal and Replicate are listed under Images, video and voice. They are for media, not chat.
  • Amazon Bedrock can be connected, but Ask AI cannot run a chat on it yet.
  • ChatGPT and Grok plans are shown as coming soon. See Use your ChatGPT or Grok plan.
  • Omne AI is listed as included with your plan, but Ask AI does not run on it yet.

Why a model may be missing

The list comes from the provider at the moment you connect. If the provider releases a model later, or your key does not have access to a model, it will not be in the menu. Opening the model menu in Ask AI reads your providers again, so a provider you connected a moment ago appears without reloading the page. To refresh a provider's model list, remove the connection and connect it again.

Choosing a provider

There is no single right answer. OpenRouter is convenient because one account reaches many models. A direct key from Anthropic, OpenAI or Google gives you that maker's billing and limits without a middle step. Cerebras is chosen for speed. Perplexity is chosen for sourced web answers. A custom address lets you point Ask AI at a server you run yourself.

Last updated November 3, 2026

Still stuck?

There's a real person behind Omne, so ask anything.

Contact support