Is OpenCode down, or is it the provider?
OpenCode runs on your machine, so "down" almost always means the model provider behind it. Zen and Go are OpenCode's own hosted models; their live status is below. "Provider is overloaded" or a retry countdown is the provider. Switch with /models, or add another provider with /connect.
Live OpenCode Zen and Go status
Checked
Degraded performance
- Console APIOperational
- Request logsOperational
- Inference API (all hosted models)Operational
- Go modelsDegraded
- Zen pay-as-you-go modelsOperational
- Free modelsDegraded
- Go models: provider errors (muse-spark-1.3-contributor)degraded
- Free models: provider errors (muse-spark-1.3-contributor-free)degraded
- Last incident: Go models: slow responses (deepseek-v4-pro)resolved Oct 15
Source: status.opencode.ai, OpenCode's hosted services; the OpenCode app itself runs on your machine. Read on our server about every 2 minutes.
Which part is down?
| What you see | What it means |
|---|---|
| "Provider is overloaded" | The provider of the current model reported overload. OpenCode retries on its own |
| "[retrying in … attempt #2]" | OpenCode is backing off after a 429, a 5xx or an overload error, with a growing wait between attempts |
| "Error from provider (…)" | Zen or Go passed on an error from upstream. The text after the colon says what failed |
| "Cannot connect to API" | An AI_APICallError: the connection to the provider failed. Check the rows above if you use Zen or Go, or that provider's own status page |
| "5-hour usage limit reached" | Or weekly, or monthly: your Go plan's window for that model, not an outage. See OpenCode rate limited |
| "Free usage exceeded" | "…subscribe to Go": the free models' limit |
| "Insufficient balance" | Your Zen balance ran out. Zen returns it as HTTP 401 |
Five minutes into an OpenCode outage
- Find which provider the current model uses: Zen, Go, or your own key. Only Zen and Go are on status.opencode.ai.
- If one model on Zen or Go is degraded above, type
/modelsand pick another model. - If Zen or Go is down as a whole, run
/connectand add another provider, then pick its model with/models. - If your own provider is failing, check that provider's status page: OpenCode only passes its errors on.
Add another provider to OpenCode
| Route | Steps |
|---|---|
| A provider OpenCode lists | Run /connect in the TUI, or opencode providers login in a shell, choose the provider and paste its key |
| Standard Compute | Run /connect, choose Standard Compute, paste your key, then pick a model with /models |
| Any OpenAI-compatible endpoint | In opencode.json, add a provider with "npm": "@ai-sdk/openai-compatible", options.baseURL and its models |
| Go over its limit | Turn on Use balance: Go then falls back to your Zen balance after its limits |
What Standard Compute does when a provider fails
- Every request goes through OpenRouter, to a provider order we review per model. If that provider fails, OpenRouter can send the request to another provider of the same model.
- Each failed call is retried twice on a 429, a 5xx or a network error, after 0.5 and then 1 second, or after the provider's Retry-After up to 8 seconds. If every attempt fails, your agent gets HTTP 503 with "The model is temporarily unavailable. Please retry shortly."
- A specific model ID always stays that model: failover changes the provider, never the model.
- OpenCode lists Standard Compute natively through models.dev, so it sits next to Zen and Go in
/connectas a second route.
The catch
A second provider only helps if it serves the model from somewhere else. When a model's maker is down at the model level, every provider of that model may be affected, Zen and Standard Compute included; pick a model from another vendor until it recovers.
Where Standard Compute fits
Standard Compute is built to make coding agents cost less: its router sends each step to the cheapest model that can do it, on providers we vet for price, uptime and speed, and a failed provider falls back automatically to another provider of the same model. An opt-in Compliance rule keeps a key, a team or the whole company on endpoints in Europe, the United States or both, and we don't train on your prompts or store them. One key at https://api.stdcmpt.com/v1 serves Claude, GPT, Gemini, DeepSeek, GLM and Kimi models, from $19 a month (USD).
FAQ
Is OpenCode down or is it my provider?
OpenCode is an app on your machine, so an outage is almost always the provider behind the model. For Zen and Go, check status.opencode.ai. For your own key, check that provider's status page. Errors like a Go usage limit or an insufficient balance are your account, not an outage.
What does 'Provider is overloaded' mean in OpenCode?
The model's provider answered with an overload error, and OpenCode is retrying with a growing wait. If it keeps failing, switch model with /models or add another provider with /connect.
What is the difference between OpenCode Zen and OpenCode Go?
Zen is OpenCode's pay-as-you-go gateway, billed per request from credits. Go is a subscription, $10 or $40 a month, with 5-hour, weekly and monthly limits per model. They are separate rows on status.opencode.ai.
Can OpenCode fall back to another provider automatically?
OpenCode retries the same model and provider. To keep working through a provider outage, connect a second provider and switch with /models, or use a gateway that fails over between providers of the same model.