Skip to content

Free tier issues: read this first, then post your report here #59387

Open

Description

@neriousy

This is the main issue for free tier problems. If you see any of these, you're in the right place:

  • 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

Please comment here instead of opening a new issue. We're closing duplicates and pointing them here so everything is in one place.

1. Make sure you're on the latest OpenCode v2

Most free tier reports come from OpenCode v1 (versions 0.x and 1.x, for example Desktop or CLI 1.19.5). v1 no longer supports free models and only gets fixes for critical bugs. Updating to v2 has fixed this for many people.

Check your version:

  • Terminal: opencode --version
  • Desktop: Settings → About

The latest v2 release is 2.2.1.

Important

The v1 Desktop app's built-in updater keeps you on 1.x. Download v2 manually from https://opencode.ai/v2/docs.

Desktop

Download v2 Desktop for macOS, Windows or Linux from https://opencode.ai/v2/docs, or on macOS:

brew install --cask opencode-desktop

Terminal (TUI / CLI)

All install options are listed at https://opencode.ai/v2/docs. The quickest on macOS and Linux:

curl -fsSL https://opencode.ai/v2/install | bash

Or use a package manager:

npm install -g @opencode/cli
bun install -g --trust @opencode/cli
brew install anomalyco/tap/opencode-v2
paru -S opencode-beta

v1 and v2 both use the opencode command. If you installed v1 with a package manager, remove it first, for example npm uninstall -g opencode-ai. Windows package managers (Chocolatey, Scoop) don't support v2. On Windows, use the Desktop app, npm, or a standalone binary.

Useful docs

2. Other things to check

  • Free models only work inside OpenCode. Other clients, proxies and direct API calls can't use them. Use a paid model or Go there instead.
  • Subscribed to Go but still seeing free limits? Run /connect, sign in with OpenCode Console in your browser (make sure you authorize the workspace with your Go subscription, or choose OpenCode Go if you use a Go API key), then pick a Go model (opencode-go/...) with /models rather than a free model (opencode/...-free).
  • Hit the real daily limit? Wait for the reset time shown, subscribe to Go, or connect another provider with /connect.

3. Still broken on the latest v2? Comment with:

  • OpenCode version, and whether you use Desktop, TUI, CLI or web
  • Operating system
  • The free model you used
  • The exact error message (a screenshot is fine)
  • When it started and the reset time you were shown
  • Whether you have Go or Console credits

Warning

Don't post email addresses, payment details or other account information here. This issue is public.

4. Still don't know what's going on?

If you're on the latest v2 and none of the above helps, or the problem is about your account, billing or Go subscription, email [email protected] with the details from step 3. Billing and account issues can't be fixed in GitHub issues.

Activity

  1. pinned this issue on Oct 26, 2026
  2. opencode-agent commented on Oct 26, 2026

    @opencode-agent
    Contributor

    Thanks for setting up this thread. Every problem listed here is decided by the hosted free tier service: usage limits, reset times, version checks and blocks on clients other than OpenCode. Reproducing any of them takes a real OpenCode account calling the free models, and this automated sandbox has no access to one. So none of them were reproduced here. Each report posted in this thread will need a maintainer with account access to look at it.

  3. leostacowskicompasso commented on Oct 26, 2026

    @leostacowskicompasso

    Here is my report:

    OpenCode v2.1.4 TUI on Windows 11 Pro 25H2

    • opencode/nemotron-3.5-lightning-free
    • opencode/mimo-v2.6-flash-free
    • opencode/deepseek-v4.1-flash-free
    • opencode/big-pickle

    You've used today's free messages. They reset at 21:00.
    Upgrade to OpenCode Go → /connect · Switch model → /models

    Image Image Image Image

    Error message started on the first time I used OpenCode today (10-26-26).
    Tried with my two Zen accounts, same errors on all models I've tried so far.
    Did not use OpenCode in two full days, was working fine on friday.

    I don't have any credits. Only Zen free-tier.

    Already tried without success:

    1. Running /connect and authenticating both my Zen accounts again;
    2. Installed OpenCode Desktop for Windows and got same behavior;
    3. Restarted machine;
    4. Tried to /connect OpenCode with API Key public.

    My current setup since migrating from v1 is pretty bare-bones, no custom plugins no custom codebase.

    All of my tracked requests on Zen's Logs have inference_failed errors with 429:

    Image

    Love this tool, would greatly appreciate any feedback/workaround for this.

  4. neriousy commented on Oct 26, 2026

    @neriousy
    MemberAuthor

    @leostacowskicompasso update to v2 please!

  5. leostacowskicompasso commented on Oct 26, 2026

    @leostacowskicompasso

    5. @leostacowskicompasso update to v2 please!

    Mentioned on my report!

    OpenCode v2.1.4 TUI on Windows 11 Pro 25H2

    Screenshot:

    Image
  6. kenjaku-dev commented on Oct 26, 2026

    @kenjaku-dev
    Image

    opencode v2.1.4
    arch lunix

  7. tomatomario45-sys commented on Oct 26, 2026

    @tomatomario45-sys
    Image

    Opencode TUI v2.1.4 on Windows 11 Pro 26H2
    Deepseek V4.1 Flash Free

    This error started yesterday at 7-8pm.

    I've tried everything, but nothing worked.

    I don't have any credits, just a free zen account.

  8. micy99 commented on Oct 27, 2026

    @micy99
    Image
    1. waiting for reset time: 6:59 AM today
    2. on 7:01 AM, just typing "cek"
    3. immediately got to wait another 24 hour.

    Opencode 2.1.4 confimed.

  9. monbooo090 commented on Oct 27, 2026

    @monbooo090

    I was previously advised to use v2, as the issue where the OpenCode Console free
    tier was exhausted after just a single request had reportedly been fixed there.

    Following this instruction, I switched to v2. While it worked as expected for
    the first day, the issue unfortunately recurred the very next day. After sending
    just one request, I was blocked again with the following message:

    You've used today's free messages. They reset at 2026/10/28 9:00.

    It appears the underlying problem has not been completely resolved in v2. Could
    you please investigate the root cause and provide a permanent fix?

    Thank you for your help.

    opencode/latest/2.1.4/desktop

  10. Kapie77 commented on Oct 27, 2026

    @Kapie77

    I hit the 24-hour limit and waited for 24 hours to pass, then I sent ONE message and got the 24-hour limit message again. This happened to me TWICE.
    It seems that if, after 24 hours, I send a message in an old session, I hit the limit again instantly; I have to create a new session to avoid this bug, but then I lose all my work.
    I’ve already tried switching models (I use the free ones) and it didn’t work; I’ve already switched computers, restarted the modem, tried resetting the sent message and sending it again, and created new sessions, but nothing works.
    I use the standard free version of OpenCode (2.1.4, desktop version for Windows) and some free models.

  11. NthanLim commented on Oct 27, 2026

    @NthanLim

    i have a same problem as @Kapie77, how to resolve this problem?

    Image
  12. JackChen890311 commented on Oct 27, 2026

    @JackChen890311

    Opencode Version: 2.1.4
    OS: MacOS 27.0.1
    Model: DeepSeek v4.1 Flash Free / MiMo v2.6 Flash Free / Big-Pickle

    This starts today and I haven't sent any message yet.
    Yesterday was fine.

    Image
  13. vaibhavpandeyprayag commented on Oct 27, 2026

    @vaibhavpandeyprayag

    13. I hit the 24-hour limit and waited for 24 hours to pass, then I sent ONE message and got the 24-hour limit message again. This happened to me TWICE.
    It seems that if, after 24 hours, I send a message in an old session, I hit the limit again instantly; I have to create a new session to avoid this bug, but then I lose all my work.
    I’ve already tried switching models (I use the free ones) and it didn’t work; I’ve already switched computers, restarted the modem, tried resetting the sent message and sending it again, and created new sessions, but nothing works.
    I use the standard free version of OpenCode (2.1.4, desktop version for Windows) and some free models.

    Same thing happened with me, after 24 hours, sending single request hits the limit again (message of reset after 24 hr)

  14. bilibili-CLY commented on Oct 27, 2026

    @bilibili-CLY

    Same problem as @leostacowskicompasso :
    Opencode v2.1.4 TUI on macOs 26.2. Every Zen Free Model returns the same message. First time found out on 10-25-26 at 10pm. No credits on my account. Have tried the methods above.
    Image

  15. 534 remaining items

  16. mhollerb commented on Nov 12, 2026

    @mhollerb

    Follow-up to #59387 (comment): the issue also reproduces with stock OpenCode, without our integrations.

    On November 12, 16:34–16:36 UTC, we tested the official OpenCode 2.2.5 run --standalone with a fresh HOME, configuration, database and cache. Production state and /etc/opencode were hidden in a temporary process namespace. No gateway, OpenEpoch, custom tools, workspace instructions, credentials or paid fallback were involved.

    Using identical report prompts and synthetic evidence (approximately 25 KB, 15 lines), with matching evidence-read outputs verified by hash:

    • A fresh unseeded report failed with HTTP 429 after a native file read.
    • Adding the joke prefix in the same message also failed with HTTP 429.
    • A separate joke turn followed by the same report succeeded and returned valid JSON.

    The failures indicated approximately 7 hours 25 minutes until reset, but the separate-turn test succeeded immediately afterward. This also shows that the single-message prefix is not a reliable workaround.

    This is a small sample, not proof of the underlying cause or a guaranteed workaround. It does establish that the failure is not confined to our custom integration. The tests used synthetic data and made no production changes.

  17. marked rate limit bug #63568 as a duplicate of this issue on Nov 12, 2026
  18. marked simple use #63617 as a duplicate of this issue on Nov 12, 2026
  19. valledemena commented on Nov 13, 2026

    @valledemena

    I've been watching this thread for weeks and the pattern is clear:

    • Many users report "You've used today's free messages" even when they haven't consumed anything, or get the same error minutes after reset.
    • Limits appear "inconsistent" (sometimes session/ID dependent, sometimes model dependent) and current docs don't show the daily allowance or reset window.
    • These drops interrupt work mid-session, causing frustration and migration to other platforms.

    My proposal (issue #63368) offers a concrete alternative:

    1. One permanent, non-frontier "backbone" model that's always available (not marked "free for a limited time").
    2. Soft limits instead of hard 24h cutoffs: a rolling window (e.g., 5h like Anthropic) or a daily budget that degrades speed/quality rather than blocking completely.
    3. Full documentation of limits per model (allowance size, reset, scope) — this directly addresses issue [FEATURE]: Document free-tier accounting for free models (idle time, pool scope, remaining allowance) #57540.

    This design especially helps open-source developers who don't pay now but may become subscribers when their projects grow; it gives them a stable, predictable experience without invisible walls.

    If you want OpenCode to be your go-to AI agent — not just an experiment you abandon when the free tier blocks you — vote with 👍 on #63368. The bot has already indexed it with explicit maintainer review; with community pull we can turn proposal into policy.


    He estado siguiendo este hilo durante varias semanas y el patrón es claro:

    • Muchos usuarios reportan el mensaje "You've used today's free messages" aun cuando no han consumido nada o vuelven a recibir el mismo error pocos minutos después del reset.
    • Los límites aparecen "inconsistentes" (a veces dependen de la sesión / ID, a veces del modelo) y la documentación actual no muestra ni el presupuesto diario ni la ventana de reset.
    • Estas caídas interrumpen el trabajo en medio de una sesión, lo que lleva a la frustración y a la migración a otras plataformas.

    Mi propuesta (issue #63368) ofrece una solución concreta:

    1. Un modelo "backbone" permanente y no-frontier que esté siempre disponible (no marcado como "free for a limited time").
    2. Límites suaves en lugar de cortes duros de 24h: una ventana móvil (por ejemplo 5h como la de Anthropic) o un presupuesto diario que, al agotarse, degrade la velocidad/precisión en vez de bloquear completamente.
    3. Documentación completa de los límites por modelo (tamaño del allowance, reset, scope) -- esto cubre directamente el issue [FEATURE]: Document free-tier accounting for free models (idle time, pool scope, remaining allowance) #57540.

    Esta arquitectura beneficia especialmente a los desarrolladores de código abierto que no pagan ahora pero podrían convertirse en suscriptores cuando su proyecto crezca; les brinda una experiencia estable y predecible sin penalizarlos con muros invisibles.

    Si quieres que OpenCode sea la herramienta de IA que uses de forma continua -- y no solo un experimento que abandonas cuando el free tier te bloquea -- vota con 👍 en #63368. El bot ya lo ha indexado y lo ha etiquetado para revisión de los mantenedores; con el apoyo de la comunidad podemos pasar de "propuesta" a "política".

  20. neekimyong commented on Nov 13, 2026

    @neekimyong

    I was only able to use it for 1 hour. If this is how it’s going to be, shouldn’t the notice about the 24-hour restriction be removed?

    Ekran görüntüsü 2026-11-13 054630 Ekran görüntüsü 2026-11-13 055035
  21. VariableVince commented on Nov 13, 2026

    @VariableVince

    Could this issue be one of the reasons for the free tier problems? Interrupted sessions silently continue in the background. Unnoticed until you see output somewhere like a github comment. If you have many old sessions with free models, do they just immediately eat all the allowed free usage somehow? Please beware: i have NOT seen this myself, not in the Console at least there was no usage that i didn't know the source of. But it may be part of the problem.

    Still waiting for a fix. My paid subscription has ended in the meantime since this and other fundamental issues (not able to use free models at all, plan mode allowing edits so agent creates and writes to files without your knowledge, MCPs and opencode-cli.exe with the Python.exe from those mcps not closing on close, restarting Interrupted sessions silently in background, etc), continue to go unsolved. It's an unusable product and all that is fixed is less important than these fundamental problems. Maintainers ignore or underestimate impact or are (even more scary) incompetent. Also how do they expect me to trust a deeply flawed product like this when it comes to things like security and safety of my data?

    Please have the better devs (the fullstack ones, the backend ones, not talking about those who only do UI or so) fix fundamental problems before getting them to continue work on less important things.

  22. suuniqo commented on Nov 14, 2026

    @suuniqo

    I have the same issue in macOS using v2:

    • Version: v2.2.6 TUI
    • OS: macOS Golden Gate 27.0.1
    • Free Model: DeepSeek V4.1 Flash Free
    • Error Message:
      You've used today's free messages. They reset at 1:50 AM.
      Upgrade to OpenCode Go → /connect · Switch model → /models
    Screenshot 2026-11-14 at 01 09 55
    • When it started: 0:50 AM, same day.
    • I don't have Go or Console credits.

    By turning off my WiFi, sending a message and getting an error, reconnecting and sending another message, I managed to fix it. However I am unable to trigger the error again, so I can't see if that works consistently.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    zenRelates to Zen

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions