Pricing
OpenCode Go Plan Limits (2026): Models, Caps and the Latest Changes
OpenCode Go plan limits: which models you get, how the dollar caps work, the new Kimi K3.1 Pro and Argon discount to Nov 30, and when Free or Console fits.
14 min read
Go plans, model list, limits and discounts checked Nov 1, 2026 on OpenCode’s Go docs; Console free models and fees checked Nov 1, 2026 on OpenCode’s Console docs; “Go usage limit exceeded” reports checked Nov 1, 2026 in OpenCode’s GitHub issues. Every source is listed at the bottom. No affiliate links on this page (how we work).
Short answer: OpenCode Go’s limits are dollar amounts on three clocks, and every model counts against them at its own rate. Go costs $10 a month, Go Plus $40, and Go Teams $20 per member. Each model has a monthly limit in dollars, and you can use at most 20% of it in any five hours and 50% in any week (verified Nov 1, 2026). Token prices are the same on every Go plan; the bigger plans raise the dollar limits.
What changed since our Oct 27 version:
- Correction: it’s not a separate wallet per model. We wrote that using up one model doesn’t touch another. The Go page added an undated line saying each model’s limit “determines how its usage counts toward those allowances”, and users report one shared Go allowance.
- Go Teams is listed: $20 per member per month; “Every member gets their own limits, $5/month higher than Go on every model”, and Extra Usage starts on.
- DeepSeek V4.1 Flash got bigger limits: $100 on Go and $200 on Go Plus, up from $80 and $160.
- Two Unlimited models, not one: Beam Free and Iceman Contributor Free are both listed as Unlimited for a limited time.
- New section on the “Go usage limit exceeded” error: what it means and what to do.
Still true from Oct 27: Kimi K3.1 Pro is on Go, and Gemini 4 Argon is 50% off with “The 50% discount applies through November 30, 2026.”
This page covers the Go tier only. For every OpenCode plan, see the OpenCode pricing tracker. We haven’t tested Go; everything below is from OpenCode’s own docs and its public GitHub issues.
Quick answer
- Price: Go $10/month and Go Plus $40/month (single-member workspaces); Go Teams $20 per member per month.
- Limit rule: “Usage limits are defined as monthly dollar amounts.” Five-hour = 20% of the monthly limit, weekly = 50%, monthly = 100%.
- Example: Kimi K3.1 Pro has a $15 monthly limit on Go ($60 on Go Plus, $20 on Go Teams). OpenCode estimates about 147 requests per five hours, 368 per week and 735 per month on Go.
- At the limit: “you can continue using the free models”, or turn on Extra Usage to fall back to your Console pay-as-you-go balance.
- Claude on Go: only Claude Haiku 5.5.
- Fine print: “Usage limits may change as we learn from early usage and feedback.”
Free vs Go vs Console at a glance
| Plan | Price | Models | Limits |
|---|---|---|---|
| Free | $0 | Free models, limited time | Not published |
| Go | $10/mo | Go model list | $ per model |
| Go Plus | $40/mo | Same as Go | Higher $ per model |
| Go Teams | $20/member/mo | Same as Go | Go + $5 per model |
| Console | Pay as you go | Paid Console models | Your balance |
Sources: Go docs and Console models, verified Nov 1, 2026. Console is OpenCode’s pay-as-you-go service, formerly called Zen.
How the Go limit actually works
OpenCode’s rule, verbatim: “Usage limits are defined as monthly dollar amounts… Each model has the following usage limits: 5-hour — 20% of the monthly limit; weekly — 50%; and monthly — 100%.” The Go page then adds: “Each model’s monthly limit below determines how its usage counts toward those allowances. Each response reports how much the request counted, as opencode.allowance.” (Go docs, verified Nov 1, 2026).
On Oct 27 we read that as one wallet per model. That reading doesn’t survive the new sentence or the reports:
- A user posted a support reply in issue #54875, shared with permission (not an OpenCode account posting): “All paid models in OpenCode Go draw from one shared allowance, with model-specific weighting. The allowances aren’t separate for each model.” (verified Nov 1, 2026)
- A desktop report in #60311 describes three workspace-wide counters, Rolling, Weekly and Monthly, rather than one set per model (verified Nov 1, 2026).
So the practical reading: there’s one Go allowance, and each model’s dollar limit sets how fast that model eats it. A model with a $15 limit burns your allowance faster per dollar of tokens than one with a $60 limit. OpenCode hasn’t published a worked example, so if exact budgeting matters, watch the opencode.allowance figure each response reports.
If you use only Kimi K3.1 Pro on Go, the three caps nest like this (our arithmetic from the docs):
| Window | Share | Go | Go Plus |
|---|---|---|---|
| 5 hours | 20% | $3.00 | $12.00 |
| Week | 50% | $7.50 | $30.00 |
| Month | 100% | $15.00 | $60.00 |
Two heavy weeks finish the month, and every other model draws on the same pot.
Monthly limits by model (Go / Go Plus / Go Teams)
As listed on the Go page, verified Nov 1, 2026:
| Model | Go | Go Plus | Go Teams |
|---|---|---|---|
| Kimi K3.1 Pro | $15 | $60 | $20 |
| Kimi K3 | $15 | $60 | $20 |
| Kimi K2.7 Code | $60 | $180 | $65 |
| GLM-5.3 | $15 | $120 | $20 |
| GLM-5.3-Flash | $60 | $180 | $65 |
| GLM-5.2 | $60 | $180 | $65 |
| Gemini 4 Argon | $30 | $120 | $35 |
| Claude Haiku 5.5 | $15 | $60 | $20 |
| GPT 6 Luna | $15 | $60 | $20 |
| GPT 5.6 Luna | $15 | $60 | $20 |
| Grok 4.7 | $15 | $60 | $20 |
| Qwen3.8 Max | $15 | $60 | $20 |
| Qwen3.8 Flash | $30 | $90 | $35 |
| Qwen3.7 Plus | $60 | $180 | $65 |
| DeepSeek V4 Pro | $15 | $60 | $20 |
| DeepSeek V4.1 Flash | $100 | $200 | $105 |
| MiniMax M3 | $60 | $180 | $65 |
| MiniMax M3.1 Flash | $20 | $80 | $25 |
| LongCat-2.0 | $60 | $240 | $65 |
| Step 5 Preview | $15 | $60 | $20 |
| MiMo-V2.6-Flash | $60 | $120 | $65 |
| MiMo-V2.6-Pro | $15 | $60 | $20 |
| MiMo-V2.5 | $60 | $120 | $65 |
| MiMo-V2.5-Pro | $15 | $60 | $20 |
| Muse Spark 1.3 | $60 | $120 | $65 |
| Muse Spark 1.2 | $60 | $120 | $65 |
| Hy4 preview | $30 | $120 | $35 |
| Hy3 | $60 | $240 | $65 |
| Beam Free | Unlimited | Unlimited | Unlimited |
| Iceman Contrib. | Unlimited | Unlimited | Unlimited |
Beam Free and Iceman Contributor Free are Unlimited only for a limited time, per the Go page. DeepSeek on Go also has peak hours, 01:00–04:00 and 06:00–10:00 UTC, Monday to Friday (6:30–9:30 AM and 11:30 AM–3:30 PM IST), per the same page.
OpenCode’s own request estimates on Go (per five hours / week / month; “actual usage varies”), verified Nov 1, 2026:
| Model | 5 hours | Week | Month |
|---|---|---|---|
| Kimi K3.1 Pro | ~147 | 368 | 735 |
| Kimi K3 | 110 | 250 | 490 |
| GLM-5.3 | 220 | 540 | 1,080 |
| Qwen3.8 Max | 160 | 400 | 810 |
| Grok 4.7 | 169 | 423 | 845 |
| Gemini 4 Argon | ~440 | 1,080 | 2,120 |
| Claude Haiku 5.5 | 3,850 | 9,620 | 19,230 |
| GPT 6 Luna | 4,230 | 10,560 | 21,130 |
| DeepSeek Flash | 43,330 | 108,330 | 216,670 |
DeepSeek Flash = DeepSeek V4.1 Flash. The estimates show why the dollar figures alone mislead: Haiku 5.5 and Kimi K3 share a $15 Go limit, but OpenCode estimates roughly 40 times more Haiku requests a month. Estimates for Go Plus, Go Teams and the other models weren’t recorded in our check; see the Go page.
What changed on Go
The Nov 1 changes (shared-allowance wording, Go Teams, DeepSeek limits) are listed at the top. Before that:
Kimi K3.1 Pro is on Go (new as of Oct 27). $2 in / $10 out / $0.20 cached read per 1M tokens (verified Oct 27, 2026).
Gemini 4 Argon is 50% off through Nov 30, 2026. Token prices are halved: $4.00 → $2.00 input, $20.00 → $10.00 output, $0.20 → $0.10 cached read per 1M (verified Oct 27, 2026). Since limits are dollars and the token price is what’s discounted, the same $30 buys roughly twice the tokens until Nov 30. That’s our read, not OpenCode’s wording.
The free extras shifted. Oct 20 listed three Unlimited models; Oct 27 listed one, Beam Free. On Nov 1 the page lists Beam Free and Iceman Contributor Free as Unlimited for a limited time. MiniMax M3.1 Flash is a paid Go model ($20 / $80), and LongCat 2.5 Preview Free is gone (LongCat-2.0 is paid, $60 / $240).
OpenCode publishes its limit rule, which is more than most. But the one sentence that tells you whether your models share a pot was added with no date and no changelog, the error doesn’t say which limit you hit, and a user had to email support and paste the answer into a GitHub issue to get a straight “shared allowance”. That’s a lot of detective work for a $10 plan, and we got it wrong on Oct 27 because of it.
What happens at the limit
Two options on the Go page (verified Nov 1, 2026):
- Keep going on free models. “If you reach the usage limit, you can continue using the free models.”
- Turn on Extra Usage. It’s optional on Go and Go Plus (on by default for Go Teams). Instead of blocking you, Go falls back to your Console pay-as-you-go balance after your limits.
If you top up Console for that, the Console docs list a processing fee of “4.6% + $0.31” per credit purchase, and budgets can “cap monthly spend” (Console models, verified Nov 1, 2026).
What “Go usage limit exceeded” means
The message is Go usage limit exceeded. People search it as “Error: Go usage limit exceeded”; in OpenCode’s GitHub issues it shows up as:
- a CLI log line
error.error="AI_APICallError: Go usage limit exceeded"(OpenCode CLI 2.0.27, #54875, verified Nov 1, 2026); - a raw response
429 {"error":{"type":"GoUsageLimitError","message":"Go usage limit exceeded"}with"metadata":{"limitName":"monthly"}(#52374, verified Nov 1, 2026), and the same with"limitName":"weekly"in #52783 (OpenCode 2.1.3, verified Nov 1, 2026).
It means your Go allowance ran out for one of the three windows: five-hour, weekly or monthly. It is not the free-tier message. The limitName field, where you can see it, tells you which window. The plain error doesn’t; the reporter in #54875 complained that it “does not say which limit was hit”. OpenCode’s repo bot replied that Go quotas are “enforced by the OpenCode Go service, not the opencode app (the CLI only shows the error it gets back)” and pointed to [email protected] (verified Nov 1, 2026).
When it resets: the five-hour window rolls, so it frees up as older usage ages out; the weekly and monthly windows take longer. In #60311 (OpenCode 2.2.1 desktop, verified Nov 1, 2026) the console showed “Rolling limit reached. Go resumes in 2h 37m.” and free models kept working. Check your Go page in the console for the exact time. OpenCode doesn’t publish fixed reset times for the weekly or monthly windows.
Your options, fastest first:
- Switch to a free model and keep working. OpenCode says free models remain available at the Go limit.
- Wait for the time the console shows if it was the five-hour window.
- Turn on Extra Usage with a Console balance and a budget, so requests past the limit bill pay-as-you-go.
- Move to Go Plus if you hit the weekly 50% cap often; same models, higher dollar limits.
- Connect your own provider with
/connectfor models you already pay for (providers docs, verified Nov 1, 2026).
One thing worth checking if the error appears on day one: in #52374 the cause was a Console credential used where a Go API key was expected, and the issue was closed as completed (verified Nov 1, 2026). Make sure you connected “OpenCode Go” with the Go key from the console.
If your error says “free usage” rather than “Go usage”, that’s a different limit: see OpenCode free usage exceeded.
Using Go outside OpenCode: OpenCode lists validated third-party clients for Go, including Claude Code and Codex (Go docs, verified Oct 27, 2026). The free-tier models are different: “Free-tier chat models require OpenCode; use paid models or Go from other clients” (Console models, verified Nov 1, 2026).
Free vs Go vs Console: which one fits
Stay on Free if you’re trying OpenCode out. The Console free models (Big Pickle plus the ones ending in “Free”) are “for a limited time”, have their own daily limit, and OpenCode doesn’t publish its size. For some, including Big Pickle, “collected data may be used to improve the model” (Console models, verified Nov 1, 2026). The full picture: Is OpenCode free?
Go pays off if you want Kimi, GLM, Argon or Qwen models daily and like a fixed $10 bill. Check the request estimates for your main model: GPT 6 Luna or Haiku 5.5 stretch a long way, Kimi K3 doesn’t.
Go Plus if you keep hitting the weekly 50% cap. Same models, same token prices, 2x to 8x higher dollar limits depending on the model.
Go Teams if several people need Go under one workspace. It’s $20 per member, with limits $5 above Go per model, and Extra Usage starts on, so set a budget.
Console instead if you use paid models Go doesn’t include, like Claude Sonnet 5.5 at $2 / $10 per 1M (Console models, verified Oct 25, 2026), or you’d rather pay only for what you use.
Your own key instead if you already pay a provider. OpenCode’s providers docs cover API keys, ChatGPT Plus or Pro sign-in, GitHub Copilot, and local models through Ollama, LM Studio and vLLM (providers docs, verified Nov 1, 2026). Go is optional.
Here’s what I’d do
Pick your main model first, then the plan. Go’s value depends on how fast your model eats the allowance, so look up its limit and estimate.
Leave Extra Usage off until you’ve hit a cap once and know your pace; then set a Console budget before turning it on. On Go Teams, set the budget on day one, because Extra Usage starts on.
Try Argon before Nov 30. It’s half price on tokens until then.
Don’t count on the Unlimited models. “Limited time” means exactly that.
Not sure Go is right for you? OpenCode Go vs Zen compares the flat plan with pay-as-you-go. For how Go compares with paying Anthropic directly, see OpenCode vs Claude Code. For the same five-hour-and-weekly pattern elsewhere, Codex usage limits. The full field is in AI coding tools compared.
If OpenCode says “1.18.0 or newer is required to use the free tier”, the OpenCode 1.18 free tier error post shows how to check and upgrade.
A provider-blocked response is not the same as hitting the Go plan cap: OpenCode provider blocked the response.
FAQ
What are OpenCode Go plan limits? Each model has a monthly dollar limit, and you can use at most 20% of it per five hours and 50% per week.
What does ‘Go usage limit exceeded’ mean in OpenCode? Your OpenCode Go allowance ran out for the five-hour, weekly or monthly window; switch to free models, wait for the reset the console shows, or turn on Extra Usage.
Do OpenCode Go models have separate limits? OpenCode’s Go page says each model’s limit sets how its usage counts toward the allowances, and user reports describe one shared Go allowance.
How much is OpenCode Go Teams? Go Teams is $20 per member per month, with limits $5 a month higher than Go on every model, per OpenCode’s Go page on Nov 1, 2026.
Does OpenCode Go include Kimi? Yes, Go lists Kimi K3.1 Pro and Kimi K3 at $15 a month each and Kimi K2.7 Code at $60, as of Nov 1, 2026.
Is Gemini Argon discounted on Go? Yes, Gemini 4 Argon token prices are 50% off on Go through November 30, 2026, per OpenCode’s Go page.
Is OpenCode Go cheaper than Claude Pro? Go is $10 a month against Claude Pro’s $20 (checked Oct 26, 2026), but the only Claude model on Go is Claude Haiku 5.5.
Sources
- OpenCode, Go docs (prices, Go Teams, limit rule, allowance wording, per-model limits, estimates, DeepSeek peak hours, Argon discount, Extra Usage, free-model fallback): https://opencode.ai/v2/docs/console/go/ (verified 2026-11-01; earlier readings 2026-10-20 and 2026-10-27; Kimi K3.1 Pro and Argon token prices and third-party clients from the 2026-10-27 reading)
- OpenCode, Console models (free models, data notes, processing fee, budgets, free models require OpenCode): https://opencode.ai/v2/docs/console/models/ (verified 2026-11-01; Claude Sonnet 5.5 rate verified 2026-10-25)
- OpenCode, Providers (connect, ChatGPT sign-in, GitHub Copilot, Ollama, LM Studio, vLLM): https://opencode.ai/v2/docs/providers/ (verified 2026-11-01)
- OpenCode GitHub issue #54875, “Go usage limit exceeded” log, bot reply, user-posted support reply: https://github.com/anomalyco/opencode/issues/54875 (verified 2026-11-01)
- OpenCode GitHub issue #52374, GoUsageLimitError with limitName monthly: https://github.com/anomalyco/opencode/issues/52374 (verified 2026-11-01)
- OpenCode GitHub issue #52783, limitName weekly: https://github.com/anomalyco/opencode/issues/52783 (verified 2026-11-01)
- OpenCode GitHub issue #60311, “Go resumes in” console message and workspace counters: https://github.com/anomalyco/opencode/issues/60311 (verified 2026-11-01)
- OpenCode, Console intro (Go and Go Plus at $10 and $40): https://opencode.ai/v2/docs/console/ (verified 2026-10-25)
- Anthropic, Claude pricing (Claude Pro $20): https://claude.com/pricing (verified 2026-10-26)