If OpenCode stops mid-task with a usage-limit error, do not start by searching for "the OpenCode limit." There is no single one. What stopped you, and what will unblock you, depends on which route you are on and the exact model ID you selected. Identify that route before changing plans or adding a credit card. Bringing your own key changes who meters and bills the request; it does not remove the provider's rate limits.
| Route | Billing & metering | Limit structure | Where to check | Fallback when capped |
|---|---|---|---|---|
| Free models | OpenCode limited-time offer | Model-specific; app-only for chat | OpenCode model docs & Console | Switch model, use Go, or connect BYOK |
| OpenCode Go | Subscription | Shared allowance: 20% per 5h, 50% weekly, 100% monthly | OpenCode Console | Use free model, enable Extra Usage, switch provider, or wait |
| Console pay-as-you-go | Per-request Console credits | Credit balance + workspace/member budgets | OpenCode Console | Top up credits, raise budget, or switch route |
| BYOK / custom provider | External provider token bill | Provider's own quota & rate limits | Provider statement & API errors | Top up, switch model/provider, or wait |
Free models and OpenCode Console
Using OpenCode Console is optional. Free models listed in OpenCode's model routing docs are limited-time offers, and free-tier chat models require OpenCode itself—other clients must use paid models or an OpenCode Go subscription.
There is no universal daily quota shared across every free model. Caps can be model-specific and are subject to change. Current documentation lists LongCat 2.5 Free at 100 requests per day, for example, but that number applies specifically to LongCat 2.5 Free and should not be generalized to other free models.
Because the free lineup and terms can change, review the current data terms for each free model before running it against private repositories. Do not assume all free models train on prompts, and do not assume all of them are zero-retention.
OpenCode Go
OpenCode Go offers three optional subscription tiers:
- Go: $10/month
- Go Plus: $40/month
- Go Teams: $20/member/month, with a two-member minimum
Paid models share a single pooled allowance; on Go Teams, each member gets an individual allowance. Usage is tracked in Console across three overlapping windows:
- Five-hour limit: 20% of the monthly allowance. The window starts on first use and lasts five hours.
- Weekly limit: 50% of the monthly allowance, resetting Monday at 00:00 UTC.
- Monthly limit: 100% of the monthly allowance, resetting when a new paid billing period begins.
Included monthly usage varies by model, and switching between paid models draws from the same shared allowance. Because different models consume that pool at different rates, Go cannot be reduced to a single fixed prompt or request count.
When you hit a Go limit, you can switch to a free model or enable Extra Usage to fall back to Console credits. Go Teams starts with Extra Usage enabled, and teams can toggle it off or on in Console.
Console pay-as-you-go
OpenCode Console also supports pay-as-you-go routing funded by credits, charging per request without a Go subscription. Console is optional and includes workspace and member budget controls to cap spend. Pay-as-you-go is neither free nor unlimited: requests stop when your credit balance runs out or when a configured workspace or member budget ceiling is reached.
BYOK and custom providers
OpenCode supports connecting external providers with /connect and selecting models with /models, including custom OpenAI-compatible endpoints (OpenCode provider docs). You do not need OpenCode Console to use a separate provider.
When you use BYOK, your provider's own quota and token pricing govern every request, and its rate limits still apply. BYOK is not automatically cheaper than OpenCode Go; it moves metering and billing from OpenCode to the endpoint you configured.
To illustrate what an itemized BYOK bill looks like in practice: in one observed run on 16 November 2026, OpenCode 1.19.5 used Kimi K3.1 Pro through Wallaby to fix one small tested bug in a toy repository. The run made 6 requests, took 76 seconds, used 41,831 input tokens and 1,075 output tokens, passed all five tests, and cost $0.070. The full tool comparison has the breakdown. This is one observed run on one toy repository, not an average or a basis for monthly and latency estimates.
BYOK is useful here as an overflow route, not a default upgrade. If you only hit a five-hour cap occasionally, waiting or a free fallback may be enough. If bursty work keeps hitting the weekly or monthly allowance, moving some tasks to a separate provider can avoid upgrading the whole subscription.
How to diagnose a limit and keep working
When OpenCode returns a usage-limit error, verify the route before changing plans:
- Identify the provider prefix and exact model ID in
/modelsso you know which billing path handled the request. - Check the matching quota surface:
- OpenCode Go: Open Console and check whether you hit the 5-hour (20%), weekly (50%, resets Monday 00:00 UTC), or monthly cap.
- Free models: Confirm you are inside OpenCode itself and check that specific model's current limit and terms.
- Console pay-as-you-go: Check remaining credits alongside workspace and member budget caps.
- BYOK: Inspect the provider's usage statement and raw rate-limit response rather than relying only on OpenCode's error message.
- Choose a continuation path based on cost and urgency: switch to a free model, enable Extra Usage on Go, route to another configured provider via
/models, or wait for the current window to reset.
FAQ
Why did I hit an OpenCode Go limit even after switching models? Paid models under Go share one allowance, and included monthly usage varies by model. Switching paid models draws from the same 5-hour, weekly, and monthly pool rather than starting a separate quota.
Why does a free model fail in another client or hit a limit unexpectedly? Free-tier chat models require OpenCode itself; other clients need paid models or Go. Free limits can also be model-specific (such as 100 requests/day listed for LongCat 2.5 Free), and both the free model lineup and terms can change.
Does BYOK remove usage limits? No. BYOK skips OpenCode Console quotas, but the external provider's own token charges, spend limits, and rate limits still apply.
How can I keep working after reaching a Go cap?
Switch to a free model, enable Extra Usage to spend Console credits, switch to another connected provider in /models, or wait for the 5-hour, weekly, or monthly window to reset.
Using Kimi K3.1 Pro via BYOK
Disclosure: Wallaby sells API access to Kimi K3.1 Pro.
Wallaby is one BYOK option you can connect alongside OpenCode's built-in routes or other external providers. For OpenCode users routing to Kimi K3.1 Pro through Wallaby, current rates are $1.80 per million fresh input tokens, $0.18 per million cached input tokens, and $9.00 per million output tokens, and new accounts include $0.50 in trial credit with no card required. Read the Kimi K3 in OpenCode setup guide, review the Kimi K3.1 Pro rate card, or register for a Wallaby API key.