OpenCode Go + DeepSeek
Flash for volume, Pro for hard agents
This page exists for the exact query “opencode go deepseek”. Confirm whether your account’s Flash is 0731 (check knowledge cutoff / model card in /models). Community issues have accused some free routes of older cutoffs.
Highlights
opencode-go/deepseek-v4-flash
Docs estimate ~31,650 req/5h (63,300 with 2× promo).
opencode-go/deepseek-v4-pro
~3,450 req/5h estimate. Use when Flash fails.
Also listed
Official X: Flash on Free tier too, with tighter limits.
2026-08-31
Go privacy table: DeepSeek ZDR renewed monthly; date will move.
Go’s DeepSeek lane
Pair with the 0731/0813 model guides so this page can stay operational (ids, caps) while those pages own benches.
Why it is trending
Some GitHub issues report that context on the Go lane was cut from 1M to 200k, or that the build served was not the 0731 generation. These are community reports; OpenCode has published no statement on them. The most reliable check is to read the model object the API actually returns.
Timeline
Flash 0731 announced on Go + Free.
Ids and dollar prices per model.
Current ZDR end date in the table we fetched.
Confirmed vs caution
Confirmed
Ids, estimate table, ZDR footnote, official X.
Caution
Two things are not guaranteed: OpenCode has never promised that every Go or Free request lands on the 0731 build, and there is no one-click Claude Code integration. What you actually get is whatever the API's model object reports.
How to choose
Volume on a $10 cap
Flash. Watch the 2× footnote.
Hard agent, still on Go
Pro, knowing the 5h estimate collapses to thousands not tens of thousands.
How to use it
opencode-go/deepseek-v4-flash | deepseek-v4-pro. /connect OpenCode Go.
On QCode
OpenCode Go is not a QCode product. If you would rather call official DeepSeek directly instead of going through a subscription lane, QCode currently lists deepseek-v4-flash and deepseek-v4-pro on /models.
FAQ
Is Go Flash 0731?
Officially the new Flash was added. Verify cutoff on your account.
Free vs Go Flash?
Same family, different limits/queues/maybe ctx. Paid Go is the reliability pitch.
Pro on Go worth it?
Only for tasks Flash fails — it eats the $12/5h cap fast.
1M context?
Official Flash/Pro are 1M. Some free routes were reported reduced — community.
ZDR after Aug 31?
OpenCode's documentation describes a monthly renewal, with the current zero-data-retention agreement running through 2026-08-31. Check OpenCode's own docs for the latest renewal status.
Use official DeepSeek instead?
If you need peak/off-peak official pricing and 2500/500 concurrency, yes.
Sources
opencode.ai/docs/go/ ; official OpenCode X; DeepSeek 0731/0813 cards.
Operational page, not another bench dump
Link 0731/0813 for numbers. Keep ids and caps here.
Related
Not affiliated with OpenCode or DeepSeek.