Skip to content

[Go] Weekly usage reaches 100% unexpectedly / unclear shared usage calculation #50457

Closed as duplicate

Description

@joaoguariglia

Description

Description

I am an active OpenCode Go subscriber and I believe there may be an issue, or at least a lack of clarity, in how the Go usage limits are calculated.

My dashboard currently shows:

  • Rolling usage: 89%
  • Weekly usage: 100%
  • Monthly usage: 51%
  • Weekly reset: approximately 6 days

Once the weekly usage reached 100%, I was unable to continue using Go models.

Actual usage

I exported my usage history for the last 30 days.

During this period, my Go usage was approximately:

  • 77 requests to Muse Spark 1.3 Contributor
  • 16 requests to Grok 4.7
  • 10 requests to DeepSeek V4.1 Flash
  • 103 requests in total

All of these requests appear in the CSV with:

billing_source = go

What is unclear

The current Go documentation lists different usage allowances for individual models.

However, based on the behavior of my account, the weekly usage limit appears to be shared globally across all Go models.

Once Weekly Usage reached 100%, switching to another Go model did not provide a separate weekly allowance.

This makes it unclear whether:

  1. Each Go model has its own independent usage allowance, or
  2. All Go models consume a single shared rolling / weekly / monthly pool, with different model-specific multipliers.

Expected behavior / clarification requested

Could you please clarify exactly how OpenCode Go usage is calculated?

Specifically:

  1. How is the Weekly Usage percentage calculated?
  2. Is the weekly limit shared across all Go models?
  3. If different models have different allowances, how are those allowances converted into the shared usage percentage?
  4. Why has my account reached 100% Weekly Usage while Monthly Usage is only at 51%?
  5. Can you confirm whether the usage shown on my account is correct?

I am not asking for a generic description of the limits. I would specifically like to understand how my actual requests resulted in 100% weekly usage.

Additional information

I can provide:

  • Screenshot of the OpenCode Go usage dashboard
  • Exported usage-30d.csv

I have intentionally not included any account-sensitive information in this public issue.

usage-30d-sanitized.csv

Image Image

Plugins

No response

OpenCode version

No response

Steps to reproduce

No response

Screenshot and/or share link

No response

Operating System

No response

Terminal

No response

Activity

  1. github-actions commented on Sep 21, 2026

    @github-actions
    Contributor

    This issue might be a duplicate of several existing issues reporting the same Go usage calculation discrepancy. Please check:

    If your issue isn't covered by any of the above, feel free to keep this open and add any distinguishing details.

  2. joaoguariglia commented on Sep 22, 2026

    @joaoguariglia
    Author

    Thanks. I reviewed the related issues, especially #41391, #42985, #41206 and #43149.

    I now understand the proposed quota-normalization logic discussed there:

    normalized quota usage = visible model cost × (60 / model Usage limit)

    So, for example:

    • a model with a $60 monthly Usage allowance has a 1× factor
    • a model with a $30 allowance has a 2× factor
    • a model with a $15 allowance has a 4× factor

    I also understand why it is possible for Weekly Usage to be 100% while Monthly Usage is around 50%:

    • weekly shared limit: $30
    • monthly shared limit: $60

    Therefore, approximately $30 of normalized usage would result in:

    • Weekly: 100%
    • Monthly: 50%

    So that part of the dashboard is understandable.

    However, my actual concern remains: I cannot reconcile my exported token usage with the approximately 51% monthly quota usage shown by the server.

    My exported CSV contains only 103 Go requests:

    • 77 × Muse Spark 1.3 Contributor
    • 16 × Grok 4.7
    • 10 × DeepSeek V4.1 Flash

    All rows have:

    billing_source = go

    The CSV contains approximately:

    • 2.79M input tokens
    • 35.62M cache-read tokens
    • 200K output tokens
    • 140K reasoning tokens

    I tried recalculating the usage using the currently published model prices and the model-specific quota normalization described in the related issues.

    Even when accounting for Grok's smaller Usage allowance and therefore its higher quota multiplier, I cannot reproduce anything close to the ~51% monthly quota displayed on my account.

    Using the published pricing and current model allowances, my estimate is roughly in the ~18–20% monthly range.

    I also tried deliberately conservative assumptions against my own case, including:

    • treating Grok requests as long-context / higher-priced where applicable
    • accounting for reasoning output
    • using less favorable DeepSeek pricing
    • applying a smaller DeepSeek Usage allowance instead of the temporary promotional allowance

    Even under those assumptions, I get roughly ~34% monthly usage, still materially below the 51% shown by the server.

    So I believe there are now two separate questions:

    1. Quota normalization: I understand that model Usage values can apply different effective multipliers to the shared quota.
    2. Actual metering: The normalized usage calculated from my exported requests still does not appear to explain the quota percentage shown on my account.

    Could someone from the OpenCode team please confirm the exact server-side calculation for my usage?

    Specifically, it would help to know:

    1. What normalized quota amount did the server calculate from these 103 requests?
    2. Which pricing/Usage allowance was applied to Muse Spark 1.3 Contributor, Grok 4.7 and DeepSeek V4.1 Flash?
    3. Are cache reads, reasoning tokens, retries, failed requests, provider-side retries, or any other usage counted toward quota but omitted from the exported CSV?
    4. Was DeepSeek V4.1 Flash calculated using its promotional Usage allowance or its normal allowance?
    5. Is there any server-side usage associated with my Go account that is not present in the exported usage-30d.csv?

    The key point is that I am no longer questioning why 100% weekly can correspond to ~50% monthly.

    I am specifically questioning how my exported usage became approximately 51% monthly usage in the first place.

    If possible, please provide a worked calculation from my attached CSV so that I can compare the server-side quota calculation with the exported usage data.

  3. Tarquinen commented on Oct 20, 2026

    @Tarquinen
    Member

    We're tracking this in #42985, so I'm closing this as a duplicate. Please follow that issue for updates.

    — from 𝕺𝖕𝖊𝖓𝕮𝖔𝖉𝖊

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

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions