> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.runtype.com/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.runtype.com/_mcp/server. # Available AI models Choose AI models from multiple providers for your Flows, Agents, and Products. [Platform keys](/user-guide/getting-started/platform-keys-vs-byok) give you access to a curated selection without provider setup. Connect your own provider keys to use additional supported models and pay providers directly. New accounts include a curated set of models, with Gemini 3.8 Flash as the default model. To view the models available to your organization, open **Settings > Models**. For an overview of the settings area, see [What's in Settings?](/user-guide/settings/whats-in-settings). ## Platform keys vs. bring your own key (BYOK) Platform keys are included with every Runtype plan. They give you access to popular models without adding provider API keys. Runtype tracks usage for platform-key calls, and you can review it in [Usage data](/user-guide/settings/usage-data). BYOK lets you connect provider API keys to use supported models from those providers. Use BYOK when you need provider-specific limits, direct provider billing, or models that platform keys do not include. BYOK is available on Team plans and higher. For details, see [Billing and plans](/user-guide/settings/billing-and-plans). To connect provider keys, open **Settings > Models**, select **Providers**, and follow [Connecting AI model providers](/user-guide/integrations/connecting-ai-model-providers). ## Supported providers Runtype supports these providers: * **Platform keys:** OpenAI, Anthropic, Google, xAI, Mixlayer, and Cloudflare Workers AI, plus the Vercel AI Gateway and Cloudflare AI Gateway. You can also connect your own key for OpenAI, Anthropic, Google, xAI, Mixlayer, and Vercel AI Gateway. * **BYOK only:** Amazon Bedrock, Vertex AI, Vertex AI (Claude), Tinfoil, and custom inference endpoints. Models from other creators, such as DeepSeek, Mistral AI, Moonshot AI (Kimi), Z.ai (GLM), and Qwen, run on platform keys through these gateways and providers. To browse the catalog, open **Settings > Models**, and then click **Add model**. Search by model name or creator, and filter by model creator, capability (vision, image generation, reasoning, tools, or web search), context length, or price. Each model card shows the model's capability badges, context length, and lowest cost per 1K tokens. ## Choosing the right model Model availability changes over time. **Add model** shows the current catalog. Use these starting points when you choose a model: * **General-purpose tasks:** GPT-5.6 Sol, Claude Sonnet 5, and Gemini 3.8 Flash provide a balance of quality and speed. * **Fast, cost-sensitive tasks:** GPT-5.6 Luna, GPT-5.6 Terra, Claude Haiku 4.5, and Gemini 3.8 Flash fit high-volume or latency-sensitive Flows. * **Complex reasoning:** GPT-5.6 Sol, Claude Opus 5, and DeepSeek V4 Flash fit multi-step reasoning tasks. * **Coding and tool-heavy Agents:** Kimi K2.7 Code and GLM-5.2 fit long tool-call loops and code generation. * **Vision and images:** GPT-5.6 Sol, Claude Sonnet 5, Claude Opus 5, Claude Haiku 4.5, and Gemini 3.8 Flash support image input. If you are new to Runtype, start with a Flow and test the same prompt with different models. Compare response quality, speed, and cost before you choose one. When you save a Flow, Runtype warns you if a prompt step's model is deprecated, is not enabled for your account, or lacks a capability that the step uses. These warnings don't block the save. For each warning and its fix, see [Model advisory warnings](/user-guide/flows/flow-validation-warnings#model-advisory-warnings). ## Smart model routing Use a routed model name such as `claude-sonnet-5` or `gpt-5.6-sol` when you want Runtype to choose a provider from the providers that serve that model. You do not need a provider-specific model identifier for common cases. To choose a provider yourself, use a provider-pinned model name, such as `anthropic:claude-sonnet-5`. For how Runtype resolves each form, see [Model routing and data flow](/user-guide/settings/model-routing-and-data-flow). To add a routed model in the dashboard, follow these steps: 1. Open **Settings > Models**, and then click **Add model**. The **Add Model** dialog opens. 2. On the model card, click **Add**. To pin a provider instead, click the arrow next to **Add** on the model card, and then select a provider. The arrow appears when more than one provider serves the model. The menu shows each provider's cost per 1K tokens. A **BYOK** badge marks a provider that runs on your own key, so connect that provider before you use the model. If none of a model's providers has a platform key, **Add** and the arrow are unavailable in this dialog. This stays true after you connect your own key for that provider. ## List models from the CLI or API To list the model catalog outside the dashboard, use one of these options: * **CLI:** Run `runtype models available`. Add `--json` for machine-readable output. * **API:** Send `GET /v1/model-configs/grouped`. Add `?configured=true` to return only the models your account has configured. * **Runtype MCP server:** Call the `list_available_models` tool. The API and MCP responses also include your account's custom models, such as the models a custom inference endpoint declares. These are marked `isCustom: true` in the grouped list and `uiCategory: "custom"` in the flat list. In an organization, every member sees the models of the organization's custom inference endpoint, including members who did not add it. Each response carries a `credentialPolicy` object that states which providers can serve a model call for your account: | `mode` | Meaning | | ----------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | `byok_only` | Your organization requires its own model keys. Only the providers in `byokProviders` serve calls, and Runtype platform keys are never used. A routed model family runs only when one of those providers serves it. | | `mixed` | Models from the providers in `byokProviders` run on your keys. Models from the providers in `platformProviders` run on Runtype platform keys. | | `platform` | Your account has connected no provider keys. Models run on Runtype platform keys. | `blockedProviders` lists the providers that an organization policy refuses until a key is connected. A model whose `keyStatus` is `needs_setup` does not run. ## Next steps Continue with these guides: * [Managing AI models](/user-guide/settings/managing-ai-models): add, remove, and set default models * [Creating and editing flows](/user-guide/flows/creating-and-editing-flows): build and test a Flow with a model * [Model routing and data flow](/user-guide/settings/model-routing-and-data-flow): understand provider selection and data flow