OpenCode pricing depends on how you buy model access: Go costs $10 per month, Zen bills usage, and your own provider bills under its own terms. The open-source client does not require either paid OpenCode service. Choose the model and expected workload first, then compare the subscription, credits, and API bill that would pay for the same work.
OpenCode pricing: Go vs Zen vs your own provider
| Access route | Price structure | Choose it when |
|---|---|---|
| OpenCode Go | $10/month; selected models and usage windows | Your regular coding work fits the included models and allowances |
| OpenCode Zen | Usage credits at published model rates, plus payment fees | You want its curated model service with usage billing |
| Your own API provider | The provider's usage rates and billing terms | You need a particular model, account, or price |
| Supported subscription connection | The connected account's applicable allowance and terms | You already have access supported by that provider integration |
| Local model | Hardware and operating costs | Your machine can run a model suitable for the task |
Service terms checked September 17, 2026: OpenCode Go, Zen pricing, and provider connections. A provider's sign-in option and an API key can charge different accounts, even inside the same client.
What the $10 Go subscription buys
Go is optional and applies to a selected, changing model lineup. Its documented limits operate across five-hour, weekly, and monthly windows: 20%, 50%, and 100% of each model's allowance respectively. Models do not all receive the same allowance. Read the row for the model you intend to use, rather than borrowing the largest number from the table.
For example, a model with a $60 allowance would have $12 available within the five-hour window and $30 within the weekly window. That describes allowance accounting, not a transferable $60 cash balance. A heavy afternoon can therefore reach a short window while the monthly allowance remains partly unused.
The optional Use balance setting continues through Zen credits after Go limits are reached. Decide whether you want that fallback before starting unattended work: it changes “pause when the allowance ends” into “continue with separately paid usage.” That can be useful for deadlines, but it belongs in the budget.
Zen fees change the effective price of small top-ups
Zen's documentation lists a payment-processing charge of 4.4% plus $0.30 per transaction. With that published structure, a hypothetical $10 credit purchase adds $0.74 in fees, while a $100 purchase adds $4.70, before any applicable tax. The fixed component makes frequent small purchases more expensive per dollar of usage.
This does not mean you should prepay a large balance just for a lower percentage fee. Compare the fee saved with how much credit you expect to use. A developer spending $8 this month may prefer a small payment even when a larger purchase has a better effective rate.
For recurring work, calculate cash paid ÷ credit actually consumed. Then compare the model's token rates. This separates a cheap headline input price from the amount that leaves your card, and avoids giving unused credit the same value as completed work.
Compare an API budget on the same workload
Suppose your monthly OpenCode usage totals 2 million uncached input tokens and 150,000 output tokens. At Kunavo's current Claude Sonnet 4.6 rates of $1.20 input and $6.00 output per million, that illustrative workload has a catalog estimate of $3.30. Cache operations, separately priced tools, and extra retries are excluded. This is reproducible arithmetic, not a claim that OpenCode normally consumes that amount.
The catalog calculation is a billing floor, not a maximum charge. When the upstream service reports its charge, Kunavo bills the greater of catalog cost and upstream cost multiplied by the applicable markup; otherwise chat usage bills at catalog cost. See Kunavo's billing rules and use the actual usage record for the final comparison. Starting requires a $10 minimum prepaid top-up. That funds a balance for usage; it is neither a $10 task charge nor a monthly subscription.
Do not compare that estimate with a $10 plan unless the plan serves the model and work you need. Different models can require different numbers of attempts. Compare a completed change and its tests, not just the nominal number of tokens the cheapest option offers.
Use our OpenCode API provider comparison to compare specific buying routes and current API prices for model rates. OpenCode vs Codex covers the client choice. Switching provider and switching agent are separate purchasing decisions.
Lower OpenCode costs in the setup you already use
Start with a scoped task and a budget you can inspect. Keep repository instructions concise, avoid repeatedly pasting large logs, and let the agent read the files that matter. Switch a difficult task to a stronger model when cheaper retries stop being productive. Reset an exhausted subscription window by waiting for its documented renewal, or deliberately select a separately funded route when finishing now is worth the charge.
For a custom provider, follow the Kunavo OpenCode setup with the configuration schema for your installed version. If a model cannot be found, use the provider and model troubleshooting guide. Once connected, verify one edit, one test run, and the recorded usage before adopting the provider for a larger project.
FAQ
Is OpenCode free?
The OpenCode client is open source. Model access is separate: OpenCode Go, Zen usage credits, a supported account connection, your own API provider, or suitable local inference. Installing the client does not include unlimited paid model usage.
What is the difference between OpenCode Go and Zen?
Go is an optional $10 monthly subscription for selected models with usage allowances. Zen is a usage-based model service. The Go documentation also describes an optional Use balance setting that continues through Zen credits after an allowance is exhausted.
Does OpenCode Go have limits?
Yes. Go documents five-hour, weekly, and monthly windows, with allowances that vary by model. Do not read a monthly dollar allowance as an unrestricted balance that can all be spent at once.
What is the cheapest OpenCode provider?
There is no provider that is always cheapest for every model and task. Compare the exact model, input and output rates, cache billing, payment fees, and completed-task cost. Go can be attractive when its models and quotas fit; a usage-based provider can be preferable for occasional work or a model outside the plan.