“Free usage exceeded, subscribe to Go” in OpenCode: an OpenCode Go alternative with no subscription
This message means you have used up the free OpenCode Zen models, and OpenCode is offering its OpenCode Go plan at $10 a month (opencode.ai/docs/go, 9 Nov 2026). From Russia that plan is hard to pay for, and Go doesn’t include Claude Sonnet, Claude Opus or GPT-6.1. The alternative is to connect your own provider in the same OpenCode. With RuAPI the agent works without a VPN, one key gives you DeepSeek, GLM, Kimi, Grok, Claude and GPT, and you pay in USDT only for the tokens you use — no subscription and no 5-hour or weekly limits.
Updated:
{
"provider": {
"anthropic": {
"options": {
"baseURL": "https://www.ruapi.ai/v1",
"apiKey": "sk-..."
}
}
},
"model": "anthropic/claude-sonnet-5-5"
}A fragment of ~/.config/opencode/opencode.json: Claude goes through the built-in anthropic provider, other models through a custom ruapi provider. Replace sk-... with your RuAPI key. You don’t need a Go subscription for this, and if you have one, it can stay alongside.
Why “subscribe to Go” appears, and why the plan is awkward from Russia
Free Zen models run out without warning.
Big Pickle, DeepSeek Flash Free, GLM-5.3-Flash Free and the other free Zen models are free only for a limited time and with a limit; the docs don’t state the size of that limit, and some of these models are blocked in certain countries — Inkling 2 Small Free, for one, in Russia (opencode.ai/docs/zen, 9 Nov 2026). According to GitHub reports, the error can appear after a couple of requests (github.com/anomalyco/opencode, issues #40078, #42128).
The Go plan is hard to pay for from Russia.
Go is $10 a month, Go Plus is $40, Go Teams is $20 per member. You subscribe in the OpenCode Console and enter billing details (opencode.ai/docs/go, 9 Nov 2026). You pay in dollars; the docs don’t list payment methods — usually that means a foreign card, and Russian bank cards generally don’t work with foreign services. Beyond the limits, Go can charge a separate Console balance, and each top-up of it carries a 4.6% + $0.31 fee (opencode.ai/docs/zen).
Go limits are counted per model.
Each model has its own monthly dollar budget: no more than 20% of it in 5 hours and 50% in a week. On Go that’s, for example, $15 a month for GLM-5.3, Kimi K3.1 Pro or Grok 4.7 and $30 for Gemini 4 Argon (until 30 Nov 2026; opencode.ai/docs/go, 9 Nov 2026). Claude Sonnet, Claude Opus and GPT-6.1 aren’t on the Go list at all. With RuAPI the only limit is your balance, and you switch models with /models.
Connecting RuAPI instead of a Go subscription: 4 steps
OpenCode stays the same; only the model provider changes. It takes about five minutes.
Sign up for RuAPI and top up
Sign in at www.ruapi.ai with email, Google, Yandex or Telegram. Top up in USDT from 5 USDT (TRC20, BEP20, TON, Polygon); USDC works too, and no card is needed. To buy USDT with rubles via SBP, see docs.ruapi.ai/en/buy-usdt. Foreign-issued cards are accepted too.
A new account usually gets $0.05 right away, so you can try OpenCode on deepseek-flash, gpt-6-luna or claude-haiku-5-5 before paying. All other models unlock after your first top-up.
Create a key and give it a budget
Console → API Keys → “Create API key”. The agent decides how many requests to make, so it’s handy to cap the key: in the “Credit limit” block turn off “Unlimited” and enter an amount. This replaces the 5-hour and weekly limits — except you set the cap yourself.
Add RuAPI to OpenCode
- The easy way is one command: in the Console click “Connect Claude Code, Cline & more”, choose OpenCode and “Copy command”. The command works once; paste it into a terminal (PowerShell on Windows): it installs OpenCode if it’s missing, writes the RuAPI address and key into the settings and opens the app.
- Manually: add a provider block to ~/.config/opencode/opencode.json as in the example at the top of the page. Claude connects through the built-in anthropic provider, other models through a provider with the npm package @ai-sdk/openai-compatible; both use baseURL https://www.ruapi.ai/v1.
- In the OpenCode V2 app the “Custom provider” window may not work yet — in that case use the file, which is supported.
Pick a model from the RuAPI group, not OpenCode Zen
- Type /models and pick a model from the “RuAPI · Claude” or “RuAPI” group, for example ruapi/deepseek-flash or anthropic/claude-sonnet-5-5. Model names look alike across groups, and the “subscribe to Go” error applies only to OpenCode Zen models.
- To avoid picking a free Zen model by accident, start OpenCode with OPENCODE_DISABLE_FREE_MODELS=1 — this hides the free models (opencode.ai/docs/zen).
- Every request and its price shows up in the RuAPI Console → “Usage Logs”.
Full guide with troubleshooting: docs.ruapi.ai/en/integrations/opencode · addresses and keys: docs.ruapi.ai/en/quickstart · prompt caching: docs.ruapi.ai/en/prompt-caching
Models from OpenCode Go — and their names on RuAPI
In both Go and RuAPI
deepseek-flash, glm-5.3
kimi-k3.1-pro, MiniMax-M3.1-Flash
qwen3.8-max, grok-4.7
gpt-6-luna, claude-haiku-5-5, gemini-4-argon
Not in Go, available on RuAPI
claude-sonnet-5-5
claude-opus-5-5
gpt-5.4 and GPT-6.1 models
exact names on the pricing page
Instead of Go limits
one balance for every model
no 5-hour or weekly limits
a budget per key
USDT from 5 USDT, balance never expires
What models cost in OpenCode via RuAPI
Prices per 1M tokens in USD — the same as on the RuAPI pricing page. No monthly fee: you’re charged only for what the agent uses.
| model | input $/M | output $/M |
|---|---|---|
| deepseek-flash | $0.17 cache read $0.0120 | $1.44 |
| gpt-6-luna | $0.12 cache read $0.0120 >272K: $0.24 | $0.60 >272K: $0.90 |
| MiniMax-M3.1-Flash | $0.18 cache read $0.0180 | $0.72 |
| glm-5.3 | $1.44 cache read $0.36 | $5.02 |
| kimi-k3.1-pro | $2.40 cache read $0.24 | $12.00 |
| grok-4.7 | $2.40 cache read $0.60 >200K: $4.80 | $7.20 >200K: $14.40 |
| claude-sonnet-5-5 | $2.40 cache read $0.12 | $12.00 |
| claude-opus-5-5 | $4.80 cache read $0.24 | $24.00 |
A fair comparison: if you work a lot on the cheap models from the Go list and have a card that works with foreign services, Go is good value — for $10 a month it gives a $100 budget on DeepSeek Flash (opencode.ai/docs/go, 9 Nov 2026). RuAPI wins when your workload is small or uneven, when you need Claude Sonnet, Opus or GPT-6.1, and when you have no card to pay with. Example: an agent step with 30,000 input tokens (24,000 of them from cache) and 1,000 output tokens costs about $0.0027 on deepseek-flash, so $10 covers roughly 3,600 such steps. Reasoning tokens are billed as output.
The grey “>272K” row is the vendor’s rate for long prompts: if a request exceeds the threshold, the whole request is billed at that rate.
OpenCode Go vs. OpenCode with RuAPI
| OpenCode + RuAPI | OpenCode Go | |
|---|---|---|
| How you pay | Pay as you go: top up and spend on tokens, no subscription | $10 a month (Go Plus $40); beyond the limits, from the Console balance if extra usage is on |
| Paying from Russia | USDT from 5 USDT, no card needed; foreign-issued cards accepted too | Billing details in the OpenCode Console; Russian bank cards usually don’t work with foreign services |
| Limits | Only your balance and the key budget | A monthly budget per model; up to 20% in 5 hours, up to 50% in a week |
| Claude and GPT | Claude Haiku, Sonnet, Opus; GPT-6 Luna, GPT-5.4, GPT-6.1 | Only Claude Haiku 5.5, GPT 6 Luna and GPT 5.6 Luna |
| Better value when | You need strong models, or your workload is small and uneven | You do a lot of work on cheap Go models and have a working foreign card |
OpenCode Go and Zen details are from opencode.ai/docs/go and opencode.ai/docs/zen (checked 9 Nov 2026). The Go model list may change.