Cerebras

Cerebras provides high-speed OpenAI-compatible inference on custom inference hardware. The plugin discovers native model metadata and pricing, with a bundled catalog for offline fallback.

Install plugin

openclaw plugins install @openclaw/cerebras-provider

Installation applies to a running Gateway automatically; otherwise it takes effect on the next startup. See Apply changes and inspect.

Getting started

openclaw onboard --auth-choice cerebras-api-key

openclaw onboard --non-interactive --accept-risk --skip-health \  --auth-choice cerebras-api-key \  --cerebras-api-key "$CEREBRAS_API_KEY"

export CEREBRAS_API_KEY=csk-...

openclaw models list --provider cerebras

Non-interactive setup

openclaw onboard --non-interactive --accept-risk --skip-health \  --mode local \  --auth-choice cerebras-api-key \  --cerebras-api-key "$CEREBRAS_API_KEY"

--mode defaults to local, so this is the same run as the Direct flag command above. Run it on the Gateway host: remote-client onboarding (--mode remote) only configures the local client connection and does not set up provider credentials on the server.

Discovery and pricing

When Cerebras auth is configured and the inference base URL is the official https://api.cerebras.ai/v1, OpenClaw reads GET /public/v1/models. This request uses public headers only: inference API keys and discovery credentials are never sent to the metadata endpoint. A custom base URL skips this public discovery rather than mixing a proxy's catalog with Cerebras metadata. Without a Cerebras credential, the runtime provider stays inactive. Public metadata listing does not establish account entitlement.

Live rows supply the native context and completion limits, reasoning and vision capabilities, and prompt/completion prices. Cerebras returns those prices as USD per-token strings; OpenClaw converts them to USD per million tokens. The public feed does not provide a separate cache tariff, so OpenClaw uses the input price for cache reads, matching Cerebras' standard input billing for cached tokens.

Successful catalogs are cached for 60 seconds. If discovery fails, returns an empty catalog, or has no usable model rows, OpenClaw uses the bundled offline seed. In the default models.mode: "merge", fresh onboarding does not copy generated model rows or prices into your config, allowing prices to refresh. Explicitly authored model rows and costs remain intact. In models.mode: "replace", discovery is disabled and onboarding keeps the offline seed as explicit config instead.

Built-in catalog

The offline fallback model has a 131,072-token context window and a 40,960-token max output. Its prices were refreshed from the native public feed on August 31, 2026.

Fresh onboarding selects GPT OSS 120B. Models retired by Cerebras are not included in the bundled catalog; existing explicit selections are not rewritten.

Prompt caching

Cerebras automatically caches matching prompt prefixes. On the native API, OpenClaw sends a stable prompt_cache_key for each conversation to improve cache routing. Cached tokens reduce prompt processing and uncached-token rate-limit usage; they are billed at the standard input token price. See Cerebras prompt caching.

Manual config

Most setups only need the API key. Use explicit models.providers.cerebras config to override model metadata in mode: "merge"; leave models empty to use discovered rows without pinning generated prices:

{  env: { vars: { CEREBRAS_API_KEY: "csk-..." } },  agents: {    defaults: {      model: { primary: "cerebras/gpt-oss-120b" },    },  },  models: {    mode: "merge",    providers: {      cerebras: {        baseUrl: "https://api.cerebras.ai/v1",        apiKey: "${CEREBRAS_API_KEY}",        api: "openai-completions",        models: [],      },    },  },}

Related