Skip to content
Subscriptions / OpenCode

OpenCode Go

$15–$60 monthly usage value depending on the model. Five-hour limit: 20% of its monthly allowance; weekly: 50%.

Published terms checked 2026-09-29

Works with

OpenCode · Claude Code · Codex · Hermes · ZCode · Pi · jcode · Kilo Code CLI

Model access

30 named models

Billing

$10 / month

What you get

$15–$60 monthly usage value depending on the model. Five-hour limit: 20% of its monthly allowance; weekly: 50%.

Allowance by model

USD usage value, not cash credit. Five-hour and weekly values are calculated from the published 20% and 50% rules. Do not sum across models. *Unlimited applies only to the named temporary free models.

GLM-5.3-Flash
5 hours$12
Week$30
Month$60
GLM-5.3
5 hours$3
Week$7.5
Month$15
GLM-5.2
5 hours$12
Week$30
Month$60
Kimi K3
5 hours$3
Week$7.5
Month$15
Kimi K2.7 Code
5 hours$12
Week$30
Month$60
Kimi K2.6
5 hours$12
Week$30
Month$60
LongCat-2.0
5 hours$12
Week$30
Month$60
LongCat 2.5 Preview Free
5 hoursUnlimited*
WeekUnlimited*
MonthUnlimited*
Official table ↗
Token accounting rates

USD per million tokens. A dash means no cache-write rate is listed, not a zero rate. DeepSeek peak: Monday–Friday 01:00–04:00 and 06:00–10:00 UTC; all other times are off-peak. Context thresholds and temporary free pricing are shown per row. These Go rates are identical on Go Plus.

GLM-5.3-Flash
Input$0.15
Output$0.50
Cache read$0.03
Cache write-
Monthly limit$60
GLM-5.3
Input$1.40
Output$4.40
Cache read$0.26
Cache write-
Monthly limit$15
GLM-5.2
Input$1.40
Output$4.40
Cache read$0.26
Cache write-
Monthly limit$60
Kimi K3
Input$3.00
Output$15.00
Cache read$0.30
Cache write-
Monthly limit$15
Kimi K2.7 Code
Input$0.95
Output$4.00
Cache read$0.19
Cache write-
Monthly limit$60
Kimi K2.6
Input$0.95
Output$4.00
Cache read$0.16
Cache write-
Monthly limit$60
LongCat-2.0
Input$0.30
Output$1.20
Cache read$0.006
Cache write-
Monthly limit$60
LongCat 2.5 Preview Free
InputFree
OutputFree
Cache readFree
Cache write-
Monthly limitUnlimited
Official table ↗
Provider request estimates

OpenCode estimates, not guaranteed request quotas. They assume specific input, cached-input and output mixes. Actual workload and cache behavior can materially change these counts. Unlimited free-model rows are temporary offers.

GLM-5.3-Flash
5 hours6,320
Week15,790
Month31,580
GLM-5.3
5 hours220
Week540
Month1,080
GLM-5.2
5 hours880
Week2,150
Month4,300
Kimi K3
5 hours110
Week250
Month490
Kimi K2.7 Code
5 hours1,350
Week3,380
Month6,750
Kimi K2.6
5 hours1,150
Week2,880
Month5,750
LongCat-2.0
5 hours11,400
Week28,600
Month57,200
LongCat 2.5 Preview Free
5 hoursUnlimited
WeekUnlimited
MonthUnlimited
Official table ↗
Token mix behind request estimates

OpenCode’s assumed token mix, not a measurement of your workload. Its request estimates depend heavily on cache reads. Use your own recorded token categories for workload comparisons.

Grok 4.7/4.6
Input / request390
Cached / request32,500
Output / request120
GLM-5.3-Flash
Input / request1,000
Cached / request55,000
Output / request200
GLM-5.3/5.2
Input / request700
Cached / request52,000
Output / request150
GPT 6 Luna
Input / request1,000
Cached / request50,000
Output / request220
GPT 5.6 Luna
Input / request1,000
Cached / request50,000
Output / request220
Kimi K3
Input / request1,050
Cached / request76,500
Output / request300
Kimi K2.7/K2.6
Input / request870
Cached / request55,000
Output / request200
LongCat-2.0
Input / request920
Cached / request88,900
Output / request200
Official table ↗
Training and data retention

Provider-published terms. DeepSeek zero-day retention depends on a monthly agreement currently valid through October 31, 2026. Muse Spark Contributor permits training and has regional restrictions. GPT Luna abuse-monitoring logs may be retained up to 30 days; Grok stored-data features affect retention.

Grok 4.7
Model trainingNot used
Retention30 days
Grok 4.6
Model trainingNot used
Retention30 days
GPT 6 Luna
Model trainingNot used
Retention30 days
GPT 5.6 Luna
Model trainingNot used
Retention30 days
GLM-5.3-Flash
Model trainingNot used
Retention0 days
GLM-5.3
Model trainingNot used
Retention0 days
GLM-5.2
Model trainingNot used
Retention0 days
Kimi K3
Model trainingNot used
Retention0 days
Official table ↗
Validated coding clients

Validation is not a guarantee for every future client release. All clients must preserve stable session identification; proxies must forward it. Copilot Chat, Kimi Code, DeepSeek Harness and MiMo Code have documented missing or incomplete support in investigated versions.

OpenCode
Session support / requirementConnect Go using /connect, then choose a model with the opencode-go/ prefix.
Hermes
Session support / requirementRequires the session-header change in PR #101864, merged after v0.21.0. Version 0.21.0 alone is insufficient.
Claude Code
Session support / requirementNative session identification is recognized; no custom-header wrapper required.
Codex
Session support / requirementNative session identification is recognized. Some versions and proxies omit it; forwarding must retain it.
ZCode
Session support / requirementIts native header is recognized. The separate x-opencode-session request is no longer required.
Pi
Session support / requirementUse a current build that sends OpenCode session information.
jcode
Session support / requirementRequires version 0.81.6 or later.
Kilo Code CLI
Session support / requirementRequires the session-header fix in PR #13752. This validation covers the CLI, not the VS Code extension.
Official table ↗
API model IDs & endpoints

These endpoints consume the Go subscription allowance. They do not establish direct-provider API pricing or a separate pay-as-you-go entitlement. Preserve the client and session headers required by Go.

Grok 4.7
API model IDgrok-4.7
Endpointhttps://opencode.ai/zen/go/v1/responses
SDK package@ai-sdk/openai
Grok 4.6
API model IDgrok-4.6
Endpointhttps://opencode.ai/zen/go/v1/responses
SDK package@ai-sdk/openai
GPT 6 Luna
API model IDgpt-6-luna
Endpointhttps://opencode.ai/zen/go/v1/responses
SDK package@ai-sdk/openai
GPT 5.6 Luna
API model IDgpt-5.6-luna
Endpointhttps://opencode.ai/zen/go/v1/responses
SDK package@ai-sdk/openai
GLM-5.3-Flash
API model IDglm-5.3-flash
Endpointhttps://opencode.ai/zen/go/v1/chat/completions
SDK package@ai-sdk/openai-compatible
GLM-5.3
API model IDglm-5.3
Endpointhttps://opencode.ai/zen/go/v1/chat/completions
SDK package@ai-sdk/openai-compatible
GLM-5.2
API model IDglm-5.2
Endpointhttps://opencode.ai/zen/go/v1/chat/completions
SDK package@ai-sdk/openai-compatible
Kimi K3
API model IDkimi-k3
Endpointhttps://opencode.ai/zen/go/v1/chat/completions
SDK package@ai-sdk/openai-compatible
Official table ↗
How usage is counted
Input, output and cached tokens consume the selected model’s dollar-denominated allowance at Go rates. These are usage accounting rates, not an extra invoice on top of the subscription. Model allowance values are not balances to add together. Source ↗
When you reach a limit
Requests are blocked unless you enable Use balance and have paid Console credits. That opt-in continues requests using pay-as-you-go credits. Free models remain available. Source ↗
Privacy matters by model
Muse Spark Contributor allows prompts and completions to be used for training and is not zero-data-retention. Grok and GPT Luna list 30-day retention. Other listed models show zero-day retention; DeepSeek’s agreement is currently valid through October 31, 2026. Source ↗
Workspace and client requirements
One Go or Go Plus subscriber per workspace. Designed for coding-agent requests. Clients need their own user agent and stable session identification. Validated clients and version restrictions are listed below. Source ↗
Regional and temporary access
Muse Spark Contributor is limited to regions allowed by Meta. LongCat 2.5 Preview Free and Space Bunny Free are temporary offers. OpenCode may change models and allowances. Source ↗
Renewal and cancellation
Recurring charges continue until you cancel through account settings. The general paid-service terms do not specify a Go-specific refund entitlement or when remaining access ends after cancellation. Source ↗
Reset and billing details still to check
The cited terms give five-hour, weekly and monthly limits but do not establish the exact reset anchor, cross-model consumption rules, rollover or a Go-specific refund policy. Check the Console before relying on those details. Source ↗

Published terms checked 2026-09-29. Provider terms can change.

Included models & access

Go and Go Plus share this model lineup, with different model-specific usage allowances. Temporary free models can change.

Included coding models30

Provider lineup ↗

Model access checked 2026-09-28. Named access does not establish how much of that model a plan includes or a predictable usage allowance.

When you reach the limit

Free models remain available. Enable Use balance with paid Console credits to continue on pay-as-you-go pricing after an included limit. Otherwise requests are blocked.

Other OpenCode subscriptions
OpenCode Go Plus$40 / month Compare ↗
Published terms, sources & history

Published price and access do not establish a numeric allowance.

  • Included usage

    Model-dependent monthly allowances, with five-hour and weekly limits.

  • Compatible tools

    OpenCode · Claude Code · Codex · Hermes

  • After the limit

    Optional Zen balance can fund usage after the included limit; free models remain available.

Catalog record opencode-go@2026-09-28. Catalog dates identify recorded rule versions. New admissions do not establish a provider launch date. Provider credits are specific to that provider. Listing a price does not establish capacity, equivalent experience or that a plan could replace another.

2026-09-28 · $10/month · checked 2026-09-28 · in effect for new subscribers