Pricing

OpenCode "Free Usage Exceeded": What the Free Limits Are and How to Fix It

Hit 'free usage exceeded' in OpenCode? What the free models allow, why the limit trips, and your options: wait, add credits, Go at $10, or bring a key.

By Engineer Foodie Desk Oct 25, 2026 · updated Nov 9, 2026 ✓ Sources checked Nov 9, 2026

13 min read

Free models, privacy notes and Console pricing checked Oct 29, 2026 in OpenCode’s Console models docs; reset behaviour and error wording checked Oct 29, 2026 in OpenCode’s pinned free-tier issue on its official GitHub repo. Go limits checked Oct 27, 2026; providers Oct 20, 2026. Other dates under Sources. No affiliate links on this page (how we work).

You were mid-session, OpenCode stopped, and the screen says “Free usage exceeded, subscribe to Go”, with a “retrying in…” countdown. Or, on newer versions, “You’ve used today’s free messages. They reset at …” with a time. Both are wordings OpenCode’s own team lists in its main free-tier issue (checked Oct 29, 2026). Some logs show a longer version: “Free usage limit reached. Subscribe to OpenCode Go to keep going: https://opencode.ai/go”.

Here’s what it means. You were on one of OpenCode’s free models (Big Pickle, or anything whose name ends in “Free”), and those models have their own usage cap. You’ve used it up. OpenCode itself still works fine. Only the free models are locked until the reset.

The fix depends on whether you want to wait, pay a little, or pay someone else. Options below, fastest first. One warning up front: adding credits does not lift the free limit, which is the single most confusing part of this whole thing.

We haven’t run OpenCode’s free tier ourselves. Everything here is from OpenCode’s docs and its official GitHub repo, quoted and dated.

When does the OpenCode free limit reset?

Daily, at the time the error shows you. OpenCode’s team calls it “the real daily limit” and says to “Wait for the reset time shown” (pinned free-tier issue, checked Oct 29, 2026). OpenCode’s docs don’t publish a fixed reset hour, a time zone or the size of the limit (Console models page and models config page, checked Oct 29, 2026).

In practice that means:

Newer OpenCode v2 builds print a clock time: “You’ve used today’s free messages. They reset at …”. Users have posted times like 8:00 p.m. and 6:00 PM in their reports, so it isn’t one global hour you can set an alarm for.
Older builds showed a “retrying in…” countdown in seconds instead. One October issue quotes “approximately 77,900 seconds”, about 21.6 hours.
Nothing on your machine resets it. OpenCode’s bot says the free usage limit and its reset “are tracked by the OpenCode service for your account, not by the app itself” (Oct 29, 2026). Reinstalling, clearing cache or switching networks won’t change the counter.

If the reset time passes and you’re still locked out, that’s a known complaint, not you doing it wrong. The pinned issue lists “the reset never happens” as a symptom. Jump to the update step below first.

How to reset or get past the limit

There’s no button that resets the free limit early. These are the real ways back to work, fastest first.

Update to the latest OpenCode v2. OpenCode’s team says “Most free tier reports come from OpenCode v1” and that v1 “no longer supports free models”. Check with opencode --version in a terminal, or Settings → About in Desktop. If it starts with 0 or 1, install v2 from opencode.ai/v2/docs; the v1 Desktop updater “keeps you on 1.x”, so download it manually (checked Oct 29, 2026). Yes, the official fix for half these reports is “have you tried turning it off and on again”, Roy from The IT Crowd style, except the “off” is uninstalling v1.
Wait for the reset time shown. Free, zero setup. If the shown reset is later the same day, fine. If it’s the next evening, that’s a workday gone, so keep reading.
Switch to a paid model. Add Console credits and pick a paid model, not a “Free” one. Credits only help once you switch. Details in the Console section below.
Connect your own provider. Run /connect and add an API key or a subscription OpenCode supports. Free models aren’t the only way to use OpenCode.
Subscribe to Go. $10 a month, with its own per-model limits. See OpenCode Go plan limits before you pay. If you already pay for Go and still see free limits, OpenCode’s fix is to run /connect, choose OpenCode Go, paste your Go API key and pick an opencode-go/... model rather than an opencode/...-free one.

Switching between free models doesn’t reliably help. One report linked from the pinned issue says all free models show the same message after switching, which fits a single account-level limit.

The OpenCode v2 wording: “Free usage limit reached” and `provider.free-limit`

OpenCode v2 doesn’t only print the friendly “You’ve used today’s free messages” box. A user who pasted the raw session record in the pinned free-tier issue (comment dated Oct 29, 2026, v2.2.1 on Fedora, checked Nov 9, 2026) shows the underlying error:

"type": "provider.free-limit",
"message": "Free usage limit reached. Connect a provider or subscribe to OpenCode Go to keep going, or switch models.",
"status": 429

The same record carries the server’s body, FreeUsageLimitError: “Free usage limit reached. Free models work in OpenCode (https://opencode.ai); subscribe to OpenCode Go to keep going: https://opencode.ai/go”. Both are the same limit as the clock-time message, a 429 from the hosted free tier. Other v2 reports in that thread quote the logs as AI.Error.QuotaExceeded: Free usage limit reached, and the Console’s Zen logs show these requests as inference_failed with a 429.

What the thread shows as of Nov 9, 2026:

Update first. OpenCode’s team lists v2.2.1 as the latest release and says v1 builds (0.x and 1.x) no longer support free models.
The reset can look stuck. Several users report that after the shown reset time passes, the next message returns the same error with a time a day later. OpenCode’s team has not published a fix or an explanation in the thread. Don’t treat it as something you misconfigured.
Workarounds from users, not from OpenCode. Some report that signing out of the OpenCode Console and back in cleared it, and others that it worked on a different network. These are single-user reports and may not work for you.
Real fix on your side: connect a provider with /connect, use a Go plan, or switch models with /models, which is what the error message itself says.

What “free usage exceeded” actually means

The clearest explanation comes from an OpenCode team member, closing an issue from a user who had $9.99 of credit and still got the error (October 2026):

“Free models (Big Pickle and the models whose names end in “Free”) have their own free usage limit and never use your Zen balance, so adding credits doesn’t lift that limit. To use your balance, pick one of the paid OpenCode Zen models. If you have OpenCode Go, select models from the OpenCode Go provider to use your plan.”

(Zen is the old name for OpenCode Console, the pay-as-you-go model gateway.) Deciding between Go and Console? OpenCode Go vs Zen compares the two paid routes.

OpenCode’s repo bot adds that “The free usage limit is applied to your account on the server” and that the Console service “counts and enforces the free-model limit, including the retry countdown” (October 2026). The app just shows what the server sends back.

What OpenCode still doesn’t publish: how big the free limit is. Neither the Console models page nor the models config docs give a number (checked Oct 29, 2026), and there’s an open request in the repo asking OpenCode to document how free-tier usage is counted.

Here’s my frustration. A free tier with an unpublished cap is fine for a demo. It’s not something you can plan a workday around. And an error that says “subscribe to Go” when your account already has paid credits is going to keep generating the same support ticket. The repo shows it does: OpenCode now closes duplicates and points them all at one pinned thread.

Other free-tier errors you might see

OpenCode’s pinned issue groups these with the main error (checked Oct 29, 2026):

Free usage exceeded, subscribe to Go
Free usage limit reached
You've used today's free messages. They reset at ... (and the reset never happens)
This version of OpenCode Desktop no longer supports free models
OpenCode's free tier can only be used from within OpenCode
OpenCode 1.18.0 or newer is required to use the free tier

The two version messages mean update. The “within OpenCode” one means you called a free model from another client, proxy or the API directly, which isn’t allowed.

If it hits on your first message

There’s a cluster of late-October 2026 reports of the error appearing on the first prompt of the day or with zero Console usage. OpenCode’s current process (checked Oct 29, 2026): update to the latest v2 first; if it still happens, comment on the pinned free-tier issue with your version, OS, the free model, the exact error and the reset time you were shown. If it’s about your account, billing or Go, email [email protected], and keep account details out of the public issue. You can’t fix a server-side counter from your terminal. Cool cool cool cool, no doubt no doubt, as Jake Peralta would say through gritted teeth.

OpenCode’s free models right now

OpenCode Console lists 10 free models, each free or available “for a limited time” (verified Oct 29, 2026): Big Pickle, GLM-5.3-Flash Free, DeepSeek V4.1 Flash Free, Beam Free, Flint 40B Free, MiMo-V2.6-Flash Free, Nemotron 3 Ultra Free, Nemotron 3.5 Lightning Free, Nemotron 3.5 Super VL Free and Jev 1.13 Free.

The list moves. LongCat 2.5 Preview Free was on it on Oct 25 and is gone now. Since Oct 20, Exo Free, MiMo-V2.5 Free, MiniMax M3.1 Flash Free and LL Alpha Free have dropped off too.

Read the privacy notes before you paste in client code:

Zero-retention, no training: GLM-5.3-Flash Free, DeepSeek V4.1 Flash Free.
No training, kept 30 days: Beam Free. It’s also not available in a list of countries the docs name.
May train on your data during the free period: Big Pickle, MiMo-V2.6-Flash Free, and Flint 40B Free (which also keeps data for 30 days).
Nemotron free endpoints: “Trial use only — do not submit personal or confidential data.”

It’s the Erlich Bachman incubator deal: free rent, and in exchange he owns a piece of what you build. Here the rent is free tokens and the piece is your prompts. Fine for side projects. Think twice for work code.

Two more rules. “Free-tier chat models require OpenCode”, so you can’t call them from another client. And if you’d rather never see them, set OPENCODE_DISABLE_FREE_MODELS=1.

The paid routes in detail

Console credits and a paid model

Console is pay-as-you-go, priced per 1M tokens, at cost plus a processing fee of “4.6% + $0.31” per credit purchase (verified Oct 29, 2026).

Cheap paid options (input / output per 1M tokens): GLM 5.3 Flash $0.15 / $0.50, DeepSeek V4.1 Flash $0.30 / $1.20, MiniMax M3 $0.30 / $1.20, GPT 6 Luna $0.10 / $0.50 up to 272K tokens. If you want Claude, Claude Sonnet 5.5 is $2 / $10 (all verified Oct 29, 2026). GLM 5.3 Flash and DeepSeek V4.1 Flash are listed as paid models next to their “Free” namesakes, so they’re the natural swap.

Set a budget first. Console “budgets cap monthly spend for the workspace and each member”, and auto-recharge “buys credits when your balance runs low”. That’s helpful or dangerous depending on whether you remember turning it on.

Go ($10) or Go Plus ($40)

This is what the error is selling. Go is $10 a month and Go Plus $40 (verified Oct 27, 2026). After subscribing, choose models from the OpenCode Go provider, not the regular Console list.

Go isn’t unlimited either. “Usage limits are defined as monthly dollar amounts”, and each model has a 5-hour limit (20% of the monthly amount), a weekly limit (50%) and a monthly limit (100%). Examples: GLM-5.3-Flash gets $60 of usage a month on Go and $180 on Go Plus; Kimi K3 gets $15 and $60 (verified Oct 27, 2026). Go is for single-member workspaces, and Claude Haiku 5.5 is its only Claude model. If you hit a Go limit, Go’s docs say you can keep going on the free models, or turn on Extra Usage to fall back to your Console balance. The full breakdown is in OpenCode Go plan limits.

Your own provider key

OpenCode is free software. It “uses the AI SDK and Models.dev to support 75+ LLM providers and it supports running local models”; you add a key with /connect (verified Oct 20, 2026). Go is “completely optional and you don’t need it to use OpenCode.”

The same docs list subscriptions that connect “with zero setup”: ChatGPT Plus, GitHub Copilot and GitLab Duo (verified Oct 20, 2026). If you already pay for one of those, this fix might cost you nothing extra. A local model costs nothing but your laptop fan. If you just want another free tier to fall back on, Google’s is worth a look: Gemini CLI free tier limits.

Free vs Go vs Go Plus vs BYO key

Option	Price	Limit	Source
Free models	$0	Daily; unsized	Docs, issues
Console credits	Per token + fee	Your budget	Console docs
Go	$10/mo	Monthly $ per model	Go docs
Go Plus	$40/mo	Higher monthly $	Go docs
BYO key	Provider’s rate	Provider’s limits	Providers docs

Free-model and Console rows verified Oct 29, 2026; Go rows Oct 27, 2026; BYO row Oct 20, 2026. Plan prices also on our OpenCode pricing tracker.

If you’re tempted to plug your Claude subscription into OpenCode, OpenCode’s own docs say no: “There are plugins that allow you to use your Claude Pro/Max models with OpenCode. Anthropic explicitly prohibits this. Previous versions of OpenCode came bundled with these plugins but that is no longer the case as of 1.3.0” (verified Oct 20, 2026). Use an Anthropic API key instead, or use Claude Code itself. See Is Claude Code free? for what that costs.

Here’s what I’d do

First, check your version. If you’re on v1, update to v2 before anything else; that’s the cheapest fix there is. If you’re on v2 and the reset time shown is later the same day, wait. If it’s the next calendar day, put a small amount of Console credit on, set a monthly budget, and switch to GLM 5.3 Flash or DeepSeek V4.1 Flash, the paid listings next to the free models you were probably on.

If you hit the wall every day, Go at $10 is the cleaner deal, and I’d only move to Go Plus after Go’s weekly cap actually bites. If you already pay for ChatGPT Plus or GitHub Copilot, connect that before you buy anything. And if the error hit on your very first message on the latest v2, post in the pinned issue or email [email protected]. Paying won’t fix a broken counter.

Comparing tools instead? See OpenCode vs Claude Code, Claude Code alternatives and AI coding tools compared.

If the message is “Provider blocked the response” rather than a usage cap, see OpenCode provider blocked the response.

FAQ

How do I reset my OpenCode limit? You can’t reset it early; wait for the reset time shown, update to the latest OpenCode v2 if the reset never comes, or switch to a paid model, your own key or Go.

When does OpenCode free usage reset? It’s a daily limit that resets at the time shown in the error; OpenCode doesn’t publish a fixed hour or time zone.

What does “Free usage exceeded, subscribe to Go” mean? You’ve used up the separate usage limit on OpenCode’s free models (Big Pickle and models ending in “Free”); switch to a paid model, Go, or your own key.

What does “You’ve used today’s free messages. They reset at” mean? It’s the newer wording of the same daily free-model limit, and the time it shows is when your free messages come back.

Does OpenCode Zen have a free limit? Yes, Zen (now OpenCode Console) free models share a server-side usage limit, while paid Zen models have no free cap and bill your balance.

Can I bypass the OpenCode free limit? No, the limit is enforced on your account on OpenCode’s server; the legit ways past it are waiting, a paid model, your own provider key or Go.

Why do I still get the error after adding credits? Credits never apply to free models; pick a paid Console model to use your balance.

Is OpenCode free? Yes, the software is free and works with your own provider key or local models, and OpenCode Console also offers limited-time free models.

What are OpenCode’s free models? As of Oct 29, 2026: Big Pickle, GLM-5.3-Flash Free, DeepSeek V4.1 Flash Free, Beam Free, Flint 40B Free, MiMo-V2.6-Flash Free, three Nemotron Free models and Jev 1.13 Free.

Is OpenCode Go worth $10? If you hit the free limit daily and want open models without per-token billing, yes; Go has monthly per-model dollar limits with 5-hour and weekly caps.

Can I use my Claude Pro or Max plan in OpenCode? No, OpenCode’s docs say Anthropic explicitly prohibits it; use an Anthropic API key instead.

Sources

OpenCode Console models (free models, privacy notes, pricing, processing fee, budgets): https://opencode.ai/v2/docs/console/models/ (verified 2026-10-29)
OpenCode models config (hiding free models): https://opencode.ai/v2/docs/models/ (verified 2026-10-29)
OpenCode GitHub issue #59387, pinned free-tier thread by an OpenCode team member (error list, daily limit, v1 vs v2, Go setup, support email): https://github.com/anomalyco/opencode/issues/59387 (verified 2026-11-09)
OpenCode GitHub issues #59687, #59983 and #60080 (reset times shown to users, limit tracked by the service): https://github.com/anomalyco/opencode/issues/59687, https://github.com/anomalyco/opencode/issues/59983 and https://github.com/anomalyco/opencode/issues/60080 (verified 2026-10-29)
OpenCode GitHub issue #45324 (error wording, team member explanation): https://github.com/anomalyco/opencode/issues/45324 (verified 2026-10-25)
OpenCode GitHub issue #58205 (log wording, free tier only inside OpenCode): https://github.com/anomalyco/opencode/issues/58205 (verified 2026-10-25)
OpenCode GitHub issues #58582 and #58655 (server-side limit, countdown): https://github.com/anomalyco/opencode/issues/58582 and https://github.com/anomalyco/opencode/issues/58655 (verified 2026-10-25)
OpenCode Console intro (Go $10 / Go Plus $40): https://opencode.ai/v2/docs/console/ (verified 2026-10-29)
OpenCode Go (limits, single-member workspaces, model list, Extra Usage): https://opencode.ai/v2/docs/console/go/ (verified 2026-10-27)
OpenCode providers (75+ providers, local models, zero-setup subscriptions, Claude Pro/Max): https://opencode.ai/docs/providers (verified 2026-10-20)