Skip to main content
Your pod’s brain is its model. Managed pods start on a curated menu you can switch between; BYOK pods run whatever slug you point them at. Either way you can change the model anytime — from chat or from Settings.

Managed tiers & their models

Each managed tier bundles a machine floor with a curated, all-included model list. Your agent starts on the tier’s default and can switch to any model in the tier — the managed proxy meters real usage per call, so switching is cost-safe, and a pod can never reach an off-tier (pricier) model.
Fast, cheap, text-only chat (no computer use). Example menu: deepseek/deepseek-chat-v3.1 (default), qwen/qwen3-235b-a22b-2507, z-ai/glm-4.5-air, openai/gpt-5-mini.
These lists are illustrative — the live per-tier menus are served from the pod catalog (GET https://compute.x402layer.cc/pods/catalog) and evolve over time. Each model in the picker shows its input/output price per 1M tokens, context window, and vision support.

Uncensored models

Some tiers include uncensored models, marked with an uncensored badge in the model picker (and on the tier card). They pick a managed model with fewer content refusals for creative or open-ended work.
Uncensored managed models still run through our upstream inference provider. Your prompts and outputs must comply with that provider’s and each model’s terms — illegal content is never permitted, and violations can get a pod suspended.

Switching models

You can change the active model without a redeploy in three ways:

/model in chat

Type /model <name> in the pod’s Chat (or on Telegram/Discord) to switch on the fly. Managed pods can switch to any model in their tier.

Chat model dropdown

Pick from the model menu in the chat UI — the same curated list, with per-model price and context on hover.

The Model setting (any OpenRouter slug)

For a durable change that applies everywhere, use Settings → Model. Enter any OpenRouter slug (provider/model, e.g. anthropic/claude-sonnet-4.5) and save.
  • Applies across chat, Telegram, and the OpenAI-compatible API — one model for every surface.
  • On managed pods it’s billed at real per-token cost; on BYOK pods it runs on your own key.
  • Saving restarts the agent (~40s) to apply it.
The Model setting accepts any valid provider/model slug — not just the tier menu — so you can pin an exact model. A Redeploy resets the active model back to the tier default; re-apply the Model setting if you want to keep a custom slug.

Bring your own key (BYOK)

A BYOK pod calls your own AI provider with your own key. You pay the provider directly; the pod bills only the machine and a small service fee.
  • Provider presets. Pick OpenRouter, OpenAI, Venice or 0G and the base URL is filled in for you. You paste a key and pick a model. Custom takes any HTTPS endpoint plus its API format: openai-completions, openai-responses, anthropic-messages or google-generative.
  • Fallbacks. Add up to 3 fallback providers. If the primary fails, the pod tries each fallback in order. Each can be a different provider with its own key.
  • Singularity Layer models. Turn on Singularity Layer models to also use our managed models from a BYOK pod, billed per call from your credits at managed prices (no included allowance). It is on by default for a BYOK ClawPod and costs nothing unless used. Pick one with /model singularity/<model-id>, or add it as a fallback.
Fallbacks and Singularity Layer models on BYOK pods are ClawPod only for now. A BYOK HermPod or Pi pod uses one provider.
From the API, set these under ai when you create a pod, and change the whole provider list later with ai.providers on update.

Deploy

Pick a tier + starting model.

OpenAI adapter

Call the agent’s model via the API.

Manage

Where the Model setting lives.