OpenCode Go is OpenCode’s monthly subscription for coding models: Go costs $10 a month and Go Plus $40 a month, according to the OpenCode Go docs. Each plan gives every model its own monthly dollar allowance, split into 5-hour, weekly and monthly limits, and you connect it inside OpenCode with an API key.
Applies to: the OpenCode Go documentation and the OpenCode Go product page as published on October 26, 2026. Last checked: October 26, 2026.
On this page
Go and Go Plus prices
The OpenCode Go docs list two monthly plans:
| Plan | Price | What OpenCode says it includes |
|---|---|---|
| Go | $10/month | Lower-cost access to the models on the Go list |
| Go Plus | $40/month | Higher usage limits across the same models |
The docs describe Go as “completely optional”: you do not need it to use OpenCode, and it works like any other model provider in OpenCode (OpenCode Go docs). The OpenCode Go page says you can cancel any time and manage the subscription in your account.
Go is not the same product as Zen. The Go page FAQ says Zen is pay-as-you-go, while Go is a monthly subscription. Zen’s billing is covered in the OpenCode Zen listing of the plans tracker.
How the usage limits work
OpenCode defines Go limits as monthly dollar amounts per model, not as a request count (OpenCode Go docs). Each model has three windows:
| Window | Share of that model’s monthly allowance |
|---|---|
| 5 hours | 20% |
| Week | 50% |
| Month | 100% |
The docs also state that token pricing is the same on Go and Go Plus; only the monthly allowance per model changes. The per-model token rates are published in the usage table on the OpenCode Go docs.
Some examples of the monthly allowance per model, from the OpenCode Go docs and the Go page:
| Model | Go monthly allowance | Go Plus monthly allowance |
|---|---|---|
| Kimi K2.7 Code | $60 | $180 |
| MiniMax M3 | $60 | $180 |
| DeepSeek V4.1 Flash | $80 | $160 |
| GPT 6 Luna | $15 | $60 |
| Claude Haiku 5.5 | $15 | $60 |
OpenCode explains the difference between models: for most models it uses bulk discounts and reserved GPU capacity to give more usage, while models it has not been able to discount get a lower included allowance (OpenCode Go docs).
Estimated requests per 5 hours
OpenCode also publishes request estimates “based on typical Go usage patterns”, with the token counts per request it assumed; actual usage varies (OpenCode Go docs). For the same models:
| Model | Go, requests per 5 hours | Go Plus, requests per 5 hours |
|---|---|---|
| Kimi K2.7 Code | 1,350 | 4,050 |
| MiniMax M3 | 3,200 | 9,600 |
| DeepSeek V4.1 Flash | 34,670 | 69,330 |
| GPT 6 Luna | 4,230 | 16,920 |
| Claude Haiku 5.5 | 3,850 | 15,380 |
The docs add that usage limits “may change as we learn from early usage and feedback” (OpenCode Go docs).
What happens at the limit
According to the OpenCode Go docs:
- You can track current usage in the OpenCode console.
- If you reach the usage limit, you can keep using the free models.
- If you also have credits on your Zen balance, you can enable the Use balance option in the console, and Go falls back to your Zen balance after the limits instead of blocking requests.
The Go page FAQ also says you can top up credit in your account if you need more usage.
Which models Go includes
The OpenCode Go docs list the models available through Go, including GLM-5.3, Kimi K3, DeepSeek V4 Pro, Qwen3.8 Max, MiniMax M3, Grok 4.7, GPT 6 Luna and Claude Haiku 5.5, and state that “the list of models may change as we test and add new ones”. The docs page and the Go page do not show identical lists on October 26, 2026, so check /models in OpenCode or the models endpoint below for the current set.
The full list with metadata is available from:
https://opencode.ai/zen/go/v1/models
Set up OpenCode Go in OpenCode
The OpenCode Go docs give three steps:
- Sign in to OpenCode Console, subscribe to Go or Go Plus, and copy your API key.
- Run
/connectin the OpenCode TUI, selectOpenCode Go, and paste your API key. - Run
/modelsin the TUI to see the models available through Go.
/connect
/models
In an OpenCode config file, a Go model ID uses the format opencode-go/<model-id>; the docs’ example for Kimi K3 is opencode-go/kimi-k3 (OpenCode Go docs).
Only one member per workspace can subscribe to Go or Go Plus (OpenCode Go docs).
Using Go with other coding agents
The Go page FAQ says you can use Go with any agent. The OpenCode Go docs narrow that down: Go is designed for OpenCode and other coding agents that send similar requests, traffic is monitored for abuse, and a client should:
- Send typical coding agent traffic.
- Identify itself with its own user agent, such as
my-coding-agent/1.0. - Send a stable session ID in
x-opencode-sessionfor each conversation.
The docs list Claude Code and Codex among the validated clients, noting that Go recognizes their native session headers (OpenCode Go docs). Go models are served from endpoints under https://opencode.ai/zen/go/v1/, with the path (chat/completions, messages or responses) depending on the model.
Privacy by model
The OpenCode Go docs publish a training and retention table per model. Most models are marked “Not used” for training. Two exceptions are the Muse Spark Contributor models, which the docs describe as discounted “in exchange for permission to use your prompts and completions to train future Meta models”. Retention is 30 days for Grok 4.7, GPT 6 Luna and Claude Haiku 5.5 and 0 days for many of the other models, so check the table for the model you pick.
Refunds and cancellation
The Go page FAQ says you can cancel any time, and that you may qualify for a refund if the charge was made within the last 14 days and you have not used your Go allowance during that billing period.
