Screenshots backing a report on anomalyco/opencode#42985 ("OpenCode Go quota usage appears ~4x higher than displayed DeepSeek V4 Flash cost").
Account: OpenCode Go ($10/month), personal workspace, subscribed 2026-10-20. opencode v2.0.22, macOS 26.6.1 (arm64).
| File | Content |
|---|---|
01-go-card-rolling-94pct.png |
Console "Go" card: Rolling 94% / Weekly 38% / Monthly 19% |
02-usage-by-model-2-rows.png |
"Usage by Model": kimi-k3 $2.72, deepseek-v4.1-flash $0.74 |
03-go-card-rolling-100pct-locked.png |
Console "Go" card: Rolling 100% + "Rolling limit reached. Go resumes in 1h 28m." / Weekly 40% / Monthly 20% |
04-usage-by-model-4-rows.png |
"Usage by Model": 4 models, $3.47 total spend |
The metered value is base x SUM(model usage / model limit) across models, so a single
$15/month-tier model (kimi-k3) can consume ~91% of the shared rolling meter and block
every paid Go model, even though real spend was $3.47 of the documented $60 monthly allowance.
No secrets are contained in these images.