OpenCode Zen vs Go: Differences and How to Choose
As of 2026-11-09, OpenCode's pay-as-you-go option lives in the OpenCode Console model catalog, often called Zen (opencode.ai/zen now opens the Console Models page): you add credits and are charged at per-million-token rates. OpenCode Go is a subscription with three monthly plans: Go at $10/month, Go Plus at $40/month and Go Teams at $20 per member per month, with limits set as monthly dollar amounts. The Console price list includes several Claude and GPT models; Go's official model list is mostly open models, and its only Claude entry is Claude Haiku 5.5 (no Opus or Sonnet). This page compares price, models, billing and limits line by line from the official docs, then covers how to choose and how to point OpenCode at your own key. (Go's official model list has 33 entries.)
Updated 2026-11-18
Switch models on the same endpoint and measure the cost per token yourself; check /models for callable models.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
Four key numbers
Go / Go Plus per month
The official docs list three monthly plans, Go, Go Plus and Go Teams ($20 per member per month): token pricing is the same for all three, and the difference is the monthly allowance per model — limits are defined as monthly dollar amounts, and token prices are per million tokens.
Console top-up fee
The Console billing docs say a processing fee of 4.6% + $0.31 is added on top of each credit purchase, so $50 of credits costs $52.61 and the full $50 goes to the balance. You pick $10, $50, $100 or another amount; usage is then charged at per-million-token rates, and auto-recharge tops the balance up at a trigger balance and amount you set.
Go limits per 5 hours / week / month
Go's limits are defined as monthly dollar amounts: the 5-hour window can use up to 20% of the monthly limit, and 50% per week. The monthly allowance differs per model — on Go, GLM-5.3-Flash is $60, Kimi K3 is $15 and DeepSeek V4.1 Flash is $100.
Go's model list
On Go's official page, Claude appears in the model list only as Claude Haiku 5.5 and open models make up most of it; the ones this page is about are Kimi K2.7 Code and GLM-5.3-Flash. The docs also say the list may change as models are tested and added.
What Zen and Go are
As of 2026-11-09, OpenCode offers two official paid options: pay-as-you-go through the OpenCode Console model catalog, often called Zen, and the Go monthly subscription. The Console docs describe the catalog as models tested and verified by the OpenCode team; its price list covers Claude, GPT, Gemini, Grok and Qwen, DeepSeek, GLM, Kimi and others, charged per request with prices listed per million tokens.
Go is mostly open coding models, plus Claude Haiku 5.5, Grok, GPT 6 Luna and Gemini 4 Argon, in three plans: Go ($10/month), Go Plus ($40/month) and Go Teams ($20 per member per month). In OpenCode both are just a provider, and the docs state that both are completely optional.
The official line (as of 2026-11-09)
On pay-as-you-go pricing, the Console docs give the goal as “Pass along price drops by selling at cost, with markup only to cover processing fees”, and the billing docs add a processing fee of 4.6% + $0.31 to each credit purchase. On Go, the docs state "Usage limits are defined as monthly dollar amounts," and token pricing is the same for Go ($10/month), Go Plus ($40/month) and Go Teams ($20 per member per month); Go Teams gives every member their own limits, $5 a month higher than Go on every model.
The changelog for OpenCode v2.2.4 (2026-10-30) says “Raised OpenCode Go limits for DeepSeek V4.1 Flash”, and the Go docs now list its monthly limit as $100 on Go and $200 on Go Plus. Earlier, the changelog for OpenCode v1.18.22 (2026-08-24) said the outdated Go first-month discount messaging and pricing were removed, and the release notes for OpenCode v1.18.32 (2026-09-21) said Grok 4.7 was added to Zen and Go; the Go docs state that the model list may change as OpenCode tests and adds new ones (“The list of models may change as we test and add new ones”). The Console docs say all models are hosted in the US and EU, and the top of the Go page now announces that Hy4 is available in Go.
Timeline
OpenCode v1.18.22 ships; the changelog says the outdated Go first-month discount messaging and pricing were removed.
OpenCode v1.18.32 was released; its release notes say Grok 4.7 was added to Zen and Go.
The official Zen and Go product and docs pages were checked; the Go docs list Go at $10/month and Go Plus at $40/month, and the only Claude model in the Go list is Claude Haiku 5.5.
Confirmed vs not listed by OpenCode
Confirmed (verbatim on the official pages)
All of the following can be checked word for word on the official OpenCode pages as of 2026-11-09: pay-as-you-go (Zen) lists prices per million tokens; you add credits of $10, $50, $100 or another amount, with a processing fee of 4.6% + $0.31 on top; auto-recharge tops up at a trigger balance and amount you set; the Console has usage tracking and budget controls; owners and admins can connect the workspace's own provider accounts.
Go is $10/month, Go Plus $40/month and Go Teams $20 per member per month; limits are defined as monthly dollar amounts, 20% per 5 hours, 50% per week, 100% per month; the Go model list includes Claude Haiku 5.5; Go and Go Plus are for single-member workspaces and can upgrade to Go Teams at any time; once the limit is used up you can keep using the free models or turn on Extra Usage on the Go page to draw on your Console balance (Go Teams starts with it on); you may qualify for a refund if you ask within 14 days of the charge and have used almost none of that period's allowance.
The Console docs also say all models are hosted in the US and EU, and the Go docs say the model list may change as models are tested and added.
Not listed, or stated inconsistently
Three things not to treat as settled: first, launch dates for Go, Go Plus and Go Teams are not listed on the official pages as of 2026-11-09; second, Go's request counts per 5 hours, week and month are OpenCode's estimates from typical usage, and the docs say "actual usage varies" and that limits may change based on early feedback; third, neither the Console model docs nor the Go docs list context lengths per model.
How to choose: Zen or Go
Pick Zen if you want Claude or GPT, or your usage varies
Pay-as-you-go (Zen) charges your Console credits for the tokens you actually use. Its price list has Claude Opus 5.5 ($4 input / $20 output), Claude Sonnet 5 ($2 / $10), GPT 6 Sol ($2 / $10 up to 272K) and more; the only Claude in Go's list is Claude Haiku 5.5, so these are not in Go. The Console adds usage tracking and budget controls, and owners and admins can connect the workspace's own provider accounts. If you use it lightly, or switch between frontier models, choose pay-as-you-go.
Pick Go if you run open models every day
Comparing the two official price lists, Kimi K3 and GLM-5.3 have the same token prices on pay-as-you-go and on Go, but the same money goes further on Go: converting from the official price list and looking at one model, Go's DeepSeek V4.1 Flash monthly limit of $100 is 10 times the $10 fee, and Kimi K3's $15 is 1.5 times.
Go Plus costs 4 times Go, but not every model's limit quadruples: Kimi K3 goes from $15 to $60 (4×), DeepSeek V4.1 Flash from $100 to $200 (2×), GLM-5.3 from $15 to $120 (8×). (monthly limits are caps converted at each model's prices, not guaranteed usage) Go and Go Plus are for workspaces with a single member; for a team, the docs offer Go Teams at $20 per member per month, where every member gets their own limits.
Connecting Zen, Go or your own provider in OpenCode
- 1Pay-as-you-go (Zen): sign in to the OpenCode Console, add your payment details and credits and copy the API key, then run /connect in OpenCode, choose the OpenCode pay-as-you-go provider and paste it; models in the config need the opencode/ prefix — the official example is opencode/
gpt-6.1-sol. - 2Go: subscribe to Go, Go Plus or Go Teams in the OpenCode Console and copy the key, then pick OpenCode Go in /connect; models need the opencode-go/ prefix — the official example is opencode-go/
kimi-k3. - 3Another service: the official docs say any provider can be repointed with the baseURL option, which is meant for proxy services or custom endpoints; keys saved by /connect live in
~/.local/share/opencode/auth.json.For an OpenAI-compatible service that is not in the /connect list, choose Other to create a provider, use the npm package @ai-sdk/openai-compatible (for /v1/chat/completions), and @ai-sdk/openai for models that use /v1/responses. - 4With QCode (following the OpenCode page in the QCode docs): in
~/.config/opencode/opencode.jsonset the baseURL of the anthropic provider tohttps://api.qcode.cc/claude/v1and the openai provider for GPT tohttps://api.qcode.cc/openai/v1; then run /connect, choose Anthropic or OpenAI, choose "Manually enter API Key", and paste your QCode key; write the model as anthropic/claude-sonnet-5or openai/gpt-6-sol.
On QCode
QCode is a multi-model API billed per token. Configure it as in step ④ above: one key can call claude-fable-5-2, claude-opus-5-5, claude-sonnet-5-5, gpt-6.2-sol and gpt-6.2-astra. Claude models go to the Anthropic-protocol endpoint, GPT goes to the OpenAI protocol, and each model's price is listed on /models. Want to try a new model for a few days first and not pay monthly? Pay-as-you-go Packs start at $4 for a one-time quota you use within 7 days (top-ups start at $5).
Frequently asked questions
What is the difference between OpenCode Zen and OpenCode Go?
The main difference is billing: Zen (on the official pages now the OpenCode Console, with pay-as-you-go pricing) charges per token, while Go is a monthly subscription. As of 2026-11-09, pay-as-you-go draws on prepaid Console credits at per-million-token rates and its price list includes Claude and GPT; Go comes as Go at $10/month, Go Plus at $40/month and Go Teams at $20 per member per month, is mostly open coding models (its only Claude is Claude Haiku 5.5), and meters usage in monthly dollar amounts across 5-hour, weekly and monthly windows.
How much is OpenCode Go? Is there still a first-month discount?
As of 2026-11-09, the official docs list Go at $10/month, Go Plus at $40/month and Go Teams at $20 per member per month, with no first-month price. The OpenCode v1.18.22 changelog (2026-08-24) says the outdated Go first-month discount messaging and pricing were removed. You can cancel any time; per the FAQ on the official Go page, you may qualify for a refund if you ask within 14 days of the charge and have used almost none of that period's allowance.
Does Go Plus give 4x the limits of Go?
Not for every model. Converting from the official price list: Go Plus at $40/month is 4 times Go; on monthly limits, Kimi K3 goes from $15 to $60 (4×), DeepSeek V4.1 Flash from $100 to $200 (2×) and GLM-5.3 from $15 to $120 (8×). OpenCode's estimated requests per 5 hours are 43,330 versus 86,670 for DeepSeek V4.1 Flash and 110 versus 440 for Kimi K3. Token prices are the same on both tiers.
What happens when my Go limit runs out?
You can keep using the models marked Free in the list: the Go docs say that once you reach the usage limit you can continue using the free models, and Crouton Alpha Free, Beam Free, Iceman Contributor Free and LongCat 2.5 Free are in that group, all marked as limited time in the Go docs (the first two also as limited regions). If you have credits in your Console balance, turn on Extra Usage on the Go page: after Go's limits are reached, requests fall back to your pay-as-you-go balance instead of being blocked, and Go Teams starts with Extra Usage on. On top of that, the allowance is counted in rolling windows: at most 20% of the monthly allowance every 5 hours and at most 50% per week.
Can I use Claude with Go?
Only one model. As of 2026-11-09, the Claude entry in the official Go model list is Claude Haiku 5.5 (monthly limit $15 on Go, $60 on Go Plus); there is no Claude Opus or Sonnet. The rest is mostly open models such as GLM, Kimi, DeepSeek, Qwen, MiniMax and MiMo, plus Grok, GPT 6 Luna and Gemini 4 Argon. To use Opus or Sonnet through OpenCode's official options, go through pay-as-you-go in the Console, often called Zen (for example, Claude Sonnet 5 is $2 input / $10 output per million tokens).
Can I skip Zen and Go and use QCode in OpenCode instead?
Yes. In OpenCode, pay-as-you-go (Zen) and Go are optional providers, and the docs state you can change any provider's address with baseURL. Per the QCode docs, set the baseURL as in the fourth step above and paste your QCode key with /connect. Billing is per token, and per-model prices are on /models.
Sources
OpenCode's Go page with its FAQ and the Console Models page that opencode.ai/zen now opens (fetched 2026-11-09: Go's three monthly plans, the free-model note, refund conditions, Console model prices); the OpenCode Console docs for Models, Billing and Providers and the Go docs (fetched the same day: the pay-as-you-go price list, credit amounts and processing fee, auto-recharge, connecting provider accounts, Go's three plans, limit rules, per-model monthly limits and estimated requests, the single-member-workspace rule and Go Teams, Extra Usage, and how to write a provider's baseURL and a custom provider); the OpenCode changelog entry for v2.2.4 and the GitHub release notes for v1.18.22 and v1.18.32. On the QCode side: the OpenCode integration and endpoints-and-API-formats pages of the QCode docs (fetched 2026-10-19).
Want to use your own key in OpenCode?
One QCode key calls claude-sonnet-5, claude-opus-5 and gpt-6-sol, billed per token, and each model's price is listed on /models.
Related reading
OpenCode Go guide
How Go's limits, model list and setup work.
OpenCode complete guide
Features and installation of the open-source terminal AI coding assistant.
OpenCode Go Free vs paid
How the free models and the Go subscription differ on limits, and when to upgrade.
Crypto payment guide
Nine coins, 56 networks, zero platform fee for AI coding API credits.
The figures and quotes on this page were checked on 2026-11-09 against OpenCode's official pages, which take precedence; changes may happen without notice. Go's request counts are OpenCode's estimates from typical usage, not a measure of your own usage. QCode is not an official OpenCode service and is not affiliated with Anomaly or OpenCode. Model availability on QCode is whatever the /models page shows.