Skip to content

OpenCode Go 5-hour limit shows ~50% consumed after only ~$1.80 of usage (DeepSeek V4 Pro, 13.9M cache reads) #43387

Closed as not planned

Description

@LongSkyTang

Bug Report

OpenCode Go 5-hour usage limit shows ~50% (≈$6 of the $12 window) consumed, but actual usage is only ~$1.80. Requests get rate-limited before the dollar limit is reached.

Actual behavior

  • 5-hour limit: ~50% consumed with < $2 real usage
  • Local opencode stats --days 1 --models output:
    • opencode-go/deepseek-v4-pro: 102 messages, 673.1K input, 143.6K output, 13.9M cache read, local cost $1.03
    • opencode-go/deepseek-v4-flash: 18 messages, 44.6K input, 4.6K output, 233.9K cache read, local cost $0.01
    • Total local cost: $1.83
  • Suspected cause: cache-read tokens billed at full input price instead of the discounted cache price ($0.022/M for DeepSeek V4 Pro), inflating usage ~30x (13.9M x $0.66 = ~$9.2)

Expected behavior

Usage should be counted at the documented Go rates (cache read = $0.022/M off-peak for V4 Pro).

Environment

  • OpenCode 1.18.18
  • Plan: OpenCode Go ($10/mo)
  • Model: opencode-go/deepseek-v4-pro, opencode-go/deepseek-v4-flash

Related: #38570, #24473, #38855, #26213

Activity

  1. github-actions commented on Aug 19, 2026

    @github-actions
    Contributor

    This issue might be a duplicate of existing issues. Please check:

    These all describe the same pattern of quota consumption appearing much higher than actual billed cost, particularly with DeepSeek models on the Go plan.

  2. Zaczero commented on Aug 19, 2026

    @Zaczero

    The $1.80 already includes the cache reads at the published rate. The 5-hour bar is that amount after Pro’s Usage factor.

    DeepSeek V4 Pro is Usage $15 in the table, so quota is visible × (60 / 15) = 4× against the shared $12 / $30 / $60 windows:

    https://opencode.ai/docs/go/#usage-limits

    50% of $12 is $6. $1.80 × 4 = $7.20. Local stats --days 1 is a calendar day; the bar is a 5-hour window, so a little of that $1.80 is already outside it.

    13.9M cache reads at $0.022/M = $0.31, which is already inside the $1.03 Pro line. At Pro input ($0.66/M) those same reads would be $9.20, and the rolling bar would already be past 75% from cache alone. It is at 50%. Cache is not being billed as input.

  3. thdxr commented on Oct 11, 2026

    @thdxr
    Member

    Thanks for the report! OpenCode 2.0 is now out and development is focused on V2, so we're closing issues filed against OpenCode 1. Please upgrade (opencode upgrade --major, details: https://anoma.ly/notes/opencode-2-0), and if this is still relevant on 2.0, reopen this issue or open a new one with your 2.0 version.

    — 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