Repository navigation
go: exhausting one model's 5h limit locks out every Go model despite documented per-model limits #54875
Closed as not planned
Description
Summary
The Go docs state usage limits are per model ("Each model has the following usage limits: 5-hour — 20% of the monthly limit"). In practice, after exhausting Qwen3.8 Max's 5-hour limit, every other model on the Go plan returned Go usage limit exceeded — including models whose own buckets were untouched — until the same 5-hour window reset.
Environment
- opencode version: 2.0.27
- OS: Windows_NT 10.0.26200 (win32 x64)
- Terminal: Unavailable: terminal environment variables are not set
- Shell: cmd.exe
- Install/channel: latest (release)
- Active plugins: none found in config
Reproduction
- Use
opencode-go/qwen3.8-maxuntil it returnsGo usage limit exceeded(~160 requests / 5h on the Go plan). - Switch the session model to a different Go model never used in that window (e.g.
opencode-go/mimo-v2.6-pro). - Send any message, including in a brand-new session.
Expected Behavior
Per the documented per-model limits, mimo-v2.6-pro should have its own independent 5-hour bucket (20% of its $15 monthly limit on Go) and should work normally.
Actual Behavior
The provider returns the identical account-level error for every Go model, including tiny background requests (small=true, title generation), across new sessions:
13:48:59 ERROR stream error providerID=opencode-go modelID=qwen3.8-max error.error="AI_APICallError: Go usage limit exceeded"
13:52:08 ERROR stream error providerID=opencode-go modelID=mimo-v2.6-pro error.error="AI_APICallError: Go usage limit exceeded"
13:52:13 ERROR stream error providerID=opencode-go modelID=mimo-v2.6-pro small=true agent=title error.error="AI_RetryError: Failed after 3 attempts. Last error: Go usage limit exceeded"
14:01:36 ERROR stream error providerID=opencode-go modelID=mimo-v2.6-pro error.error="AI_APICallError: Go usage limit exceeded"
14:02:31 ERROR stream error providerID=opencode-go modelID=mimo-v2.6-pro (new session, same error)
All timestamps UTC, 2026-10-12. No Go model worked again until the 5-hour window reset.
Additional Context
- The error message
Go usage limit exceededdoes not say which limit was hit (per-model vs. account-wide), making this very hard to diagnose from the client side. - Either enforcement is account-wide (contradicting https://opencode.ai/v2/docs/console/go), or the docs are misleading. Related community reports:
- https://www.reddit.com/r/opencode/comments/1vx13o9/opencode_gos_usage_limits_have_become_absurdly/
- https://www.reddit.com/r/opencode/comments/1vkh3eh/5h_limits_boasted_in_opencodego_website_are_fake/
- https://www.reddit.com/r/opencodeCLI/comments/1u35gzj/opencode_go_is_a_total_lie_imo/
- https://www.reddit.com/r/opencodeCLI/comments/1vbkpk6/how_do_you_manage_usage_limits_on_opencode_go/
- Suggested fixes: (a) enforce/clarify truly per-model buckets, and/or (b) include the limit scope and reset time in the error surfaced to the client.
- Reproducibility: consistently, once the 5h bucket of the active model is exhausted (observed once first-hand; matches multiple community reports).
Reacted by 57382
Activity
opencode-agent commentedon Oct 12, 2026
Thanks for the detailed report and logs. Being locked out of every Go model after one model hits its 5-hour limit, when the docs describe per-model limits, is confusing, and the error doesn't help by not saying which limit was hit.
The Go plan's quotas are enforced by the OpenCode Go service, not the opencode app (the CLI only shows the error it gets back), and billing and account support isn't handled in GitHub issues. Please email help@anoma.ly with the details: the times you were blocked, the models involved and what your Go console showed. If you've already emailed, the team will get to it, especially now that we've seen it here. Please keep payment details, email addresses, workspace IDs and other account information out of this public issue and put them only in the email.
Other people have reported the same behaviour, ordered by likelihood:
The bot will close this issue in 1 day.
redcasinotimonneo commentedon Oct 12, 2026
Hey @shayaandanishansari, please read my comment to the exact issue you are having
#49014 (comment)
Hope it helps
p.s. i am not related to opencode team
shayaandanishansari commentedon Oct 12, 2026
shayaandanishansari commentedon Oct 12, 2026
Official confirmation from OpenCode support
This issue has been confirmed by the OpenCode support team. Here is their response (shared with permission from the original reporter):
What this confirms
The per-model "monthly limits" in the Go docs are not separate budgets. They are illustrations of how much of that model you would consume if you spent the entire shared allowance on it. The docs at https://opencode.ai/v2/docs/console/go present them in a way that implies separate per-model limits, which is misleading.
Switching to a different paid model after exhausting the shared allowance does not work. This is by design, not a bug. All paid models draw from the same pool.
Free models remain available even when the shared paid allowance is exhausted.
Zen credits can be used as a bypass via the "Use balance" toggle in the console.
Impact
The Go docs should be updated to clearly state that all paid models share a single allowance, and that the per-model "monthly limit" figures are not separate budgets. The current presentation leads users to believe they can use multiple models independently within their plan, which is not the case.
opencode-agent commentedon Oct 13, 2026
Closing this now, since Go plan usage limits are handled by billing and account support rather than in GitHub issues. Thanks again for the clear write-up, @shayaandanishansari. If anything is still unresolved, please email help@anoma.ly with the details. Keep any account or payment information in the email, not here. For the wider question of per-model versus account-wide limits, see #52783.
gateway: the Go rail is one account allowance — latch_scope account, …