OpenCode Go + DeepSeek
Flash for volume, Pro for hard agents
As of 2026-11-09, the OpenCode Go docs list DeepSeek V4.1 Flash with a monthly limit of $100 on Go and $200 on Go Plus, and DeepSeek V4 Pro with $15 and $60; the estimated requests per 5 hours on Go are 43,330 for Flash and 1,050 for Pro, and the DeepSeek ZDR agreement is valid through 2026-11-30. The figures further down this page were rechecked on 2026-11-09. As of 2026-10-19, the OpenCode Go docs list two DeepSeek models: DeepSeek V4.1 Flash (model id deepseek-v4.1-flash) and DeepSeek V4 Pro (deepseek-v4-pro), written in OpenCode as opencode-go/deepseek-v4.1-flash or opencode-go/deepseek-v4-pro. The older id deepseek-v4-flash is no longer in the official Go docs or in the list returned by the Go models endpoint. All ids, prices, monthly limits, peak hours and privacy terms here come from official pages.
Updated 2026-11-10
One key for Claude, GPT and Chinese models — put what this page describes into practice.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
Highlights
opencode-go/deepseek-v4.1-flash
Monthly limit $100 on Go and $200 on Go Plus; OpenCode estimates about 43,330 requests per 5 hours (86,670 on Go Plus).
opencode-go/deepseek-v4-pro
Monthly limit $15 on Go and $60 on Go Plus; OpenCode estimates about 1,050 requests per 5 hours (4,200 on Go Plus).
Peak price is 2× off-peak
The docs define peak hours as 01:00-04:00 and 06:00-10:00 UTC, Monday through Friday; all other hours, weekends included, are off-peak. Flash costs $0.15 input / $0.60 output off-peak and $0.30 / $1.20 at peak (per million tokens).
Agreement valid through 2026-11-30
In the Go privacy table the DeepSeek rows read “Not used” for training and “0 days*” for retention; the footnote says the ZDR agreement is renewed monthly and the current one is valid through 2026-11-30. The date changes, so go by the docs at the time.
Go’s DeepSeek lane
As of 2026-11-09, OpenCode Go has two DeepSeek models, V4.1 Flash and V4 Pro, both at https://opencode.ai/inference/go/openai/v1/chat/completions with the @ai-sdk/openai-compatible package. Converting from the official price list, Pro's off-peak price is about 4.4 times Flash's for input and 3.3 times for output, and its monthly limit on Go is $15 against Flash's $100, so Flash is for volume and Pro for tasks Flash cannot handle. The Console pay-as-you-go price list, often called Zen, has both too, at different prices; see the comparison below.
What changed (as of 2026-11-09)
Several updates compared with the previous version of this page. Limits: the OpenCode v2.2.4 changelog (2026-10-30) says the Go limits for DeepSeek V4.1 Flash were raised; the Go docs now give $100 on Go and $200 on Go Plus, with about 43,330 estimated requests per 5 hours on Go. Ids: among DeepSeek models, the official Go docs list only deepseek-v4.1-flash and deepseek-v4-pro; deepseek-v4-flash is not there. Free tier: the Console model docs (the pay-as-you-go side, often called Zen) list DeepSeek V4.1 Flash Free among the free models; per the official note it is free on OpenCode for a limited time, and its provider follows a zero-retention policy and does not use data for model training. Also, the release notes for OpenCode v1.18.32 (2026-09-21) already said DeepSeek V4.1 Flash was added to the Zen docs. On Go, the peak and off-peak prices for DeepSeek are unchanged, and the ZDR agreement is still renewed monthly, currently valid through 2026-11-30. The latest OpenCode release is v2.2.6 (published 2026-11-06). DeepSeek's website announced that V4.1 Flash has been updated to the V4.1-Flash-1017 version; officially, lower cost and faster completion are claimed.
Timeline
OpenCode v1.18.32 was released; its release notes say DeepSeek V4.1 Flash was added to the Zen docs (the same release also added Grok 4.7 to Zen and Go).
Fetched opencode.ai/docs/go, the Go models endpoint and docs/zen: ids, dollar prices, monthly limits, estimated requests and the ZDR footnote for the two DeepSeek models; the DeepSeek website announcement and the latest OpenCode release (v1.19.3, published 2026-10-16) were checked the same day.
The Go privacy footnote says the current DeepSeek ZDR agreement is valid through this day; the agreement is renewed monthly, so the date after renewal is whatever the docs say at the time.
Confirmed vs caution
Confirmed
As of 2026-11-09, all of the following can be checked word for word on official pages: the model ids are deepseek-v4.1-flash and deepseek-v4-pro; the monthly limit is $100 on Go and $200 on Go Plus for Flash, and $15 and $60 for Pro; the 5-hour, weekly and monthly allowances are 20%, 50% and 100% of the monthly limit; peak hours are 01:00-04:00 and 06:00-10:00 UTC, Monday through Friday; the privacy table shows DeepSeek as not used for training with 0 days* retention; the ZDR agreement is currently valid through 2026-11-30; and the Go docs list a third-party client named DeepSeek Harness among clients with incomplete session information; that is a client compatibility issue, unrelated to the DeepSeek models themselves.
Caution
A few things the official pages do not say, so do not treat them as settled: neither the Go docs nor the Console model docs list a context length for the two DeepSeek models; the estimated request counts are computed from the typical token mix the docs give (for Flash, 410 input, 71,300 cached and 310 output tokens per request), and the docs say actual usage varies and that limits may change with early feedback; the model list can change too (“The list of models may change”).
How to choose
Volume work: Flash
Flash's monthly limit is far higher than Pro's ($100 versus $15 on Go), and OpenCode estimates about 43,330 versus 1,050 requests per 5 hours. Its off-peak price is the lowest, $0.15 input and $0.60 output per million tokens. Start with Flash for everyday volume.
Hard tasks: Pro
Pro costs more: $0.66 input / $1.98 output off-peak and $1.32 / $3.96 at peak per million tokens, with a small monthly limit and about 1,050 estimated requests per 5 hours on Go. Use it only for tasks Flash cannot handle and watch how fast the allowance drains; for a higher allowance, Pro's monthly limit on Go Plus is $60.
How to use it
① Subscribe to Go or Go Plus in the OpenCode Console and copy your API key. ② In OpenCode run /connect, choose OpenCode Go and paste the key. ③ Run /models to pick DeepSeek, or write opencode-go/deepseek-v4.1-flash (volume) or opencode-go/deepseek-v4-pro (hard tasks) in your config. ④ For an application calling the API directly, the endpoint is https://opencode.ai/inference/go/openai/v1/chat/completions with the @ai-sdk/openai-compatible package. ⑤ The docs ask clients to identify themselves with their own user agent and to send a stable session ID in x-opencode-session so routing and prompt caching can be optimized. Go and Go Plus are for single-member workspaces only.
On QCode
When OpenCode Go checkout does not work for you or its monthly limits run short, connect a per-token API in OpenCode and keep working. The DeepSeek models callable on QCode are deepseek-v4.1-flash, deepseek-v4-flash and deepseek-v4-pro. Billed per token; per-model rates are listed on /models. The configuration is in the OpenCode integration page of the QCode docs. Not ready for a monthly plan? Pay-as-you-go Packs start at $4 for a one-time quota you use within 7 days (top-ups start at $5).
FAQ
What is the model id of DeepSeek Flash on Go now?
As of 2026-10-19, the official Go docs and the Go models endpoint list deepseek-v4.1-flash (DeepSeek V4.1 Flash), written in OpenCode as opencode-go/deepseek-v4.1-flash; deepseek-v4-flash is not in the list.
What is the DeepSeek allowance on Go?
As of 2026-11-09, the monthly limit for Flash is $100 on Go and $200 on Go Plus, and for Pro $15 on Go and $60 on Go Plus. Each model's 5-hour allowance is 20% of the monthly limit, weekly 50%, monthly 100%. OpenCode estimates about 43,330 requests per 5 hours for Flash (86,670 on Go Plus) and about 1,050 for Pro (4,200 on Go Plus); actual usage varies. The OpenCode v2.2.4 changelog (2026-10-30) says the Go limits for DeepSeek V4.1 Flash were raised.
When are peak and off-peak hours, and how big is the gap?
The docs define peak as 01:00-04:00 and 06:00-10:00 UTC, Monday through Friday; all other hours, weekends included, are off-peak. Per the official price list, the peak price is 2 times the off-peak price: Flash input goes from $0.15 to $0.30 and output from $0.60 to $1.20; Pro input from $0.66 to $1.32 and output from $1.98 to $3.96 (per million tokens).
Is Pro cheaper on Go or on Zen?
It depends on how you use it. On pay-as-you-go in the Console, often called Zen, DeepSeek V4 Pro is $1.74 input / $3.48 output and V4.1 Flash is $0.30 / $1.20 per million tokens. On Go, Pro costs $0.66 input / $1.98 output off-peak, below Zen; at peak, input at $1.32 is below Zen but output at $3.96 is above Zen's $3.48. Go's advantage is the fixed monthly fee, but Pro's monthly limit is small ($15 on Go).
Is the DeepSeek ZDR agreement still valid?
As of 2026-11-09, the footnote in the Go privacy table reads: “DeepSeek: ZDR agreement is renewed monthly. The current agreement is valid through November 30, 2026.” So it is renewed monthly and currently valid through 2026-11-30. The date moves after each renewal, so go by the official docs at the time.
How else can I use these two DeepSeek models without subscribing to Go?
You can use pay-as-you-go in the OpenCode Console, often called Zen; its price list has DeepSeek V4.1 Flash and V4 Pro. You can also use QCode directly: deepseek-v4.1-flash, deepseek-v4-flash and deepseek-v4-pro are billed per token, with per-model rates on /models. Not ready for a monthly plan? Pay-as-you-go Packs start at $4 for a one-time quota you use within 7 days (top-ups start at $5).
Sources
opencode.ai/docs/go (fetched 2026-11-09: model list, per-model monthly limits and estimated requests, DeepSeek peak and off-peak prices, endpoints and AI SDK packages, the privacy table and ZDR footnote, usage requirements and clients with incomplete session support); opencode.ai/docs/zen, which now shows the OpenCode Console Models docs (same day: DeepSeek pay-as-you-go prices and the free-model list); the OpenCode changelog (same day: v2.2.4 of 2026-10-30 raised the Go limits for DeepSeek V4.1 Flash, and the latest release is v2.2.6 of 2026-11-06); opencode.ai/inference/go/v1/models and the DeepSeek website announcement (fetched 2026-10-19: the model id list; V4.1 Flash updated to V4.1-Flash-1017); the v1.18.32 release notes are a historical entry. On the QCode side: the OpenCode integration page of the QCode docs (fetched 2026-10-19; the page itself is marked updated 2026-10-15).
Want QCode's DeepSeek inside OpenCode?
One QCode key calls deepseek-v4.1-flash, claude-sonnet-5 and gpt-6-sol, billed per token.
Related
OpenCode Go
How Go's limits, model list and setup work.
GitHub Copilot BYOK for Claude and GPT
BYOK setup for VS Code Custom Endpoint and Copilot CLI: URL forms, plan requirements and the JetBrains limit.
Cursor custom API key and Base URL
Where keys go in Settings → Models, which slot can change the Base URL, why Tab skips your key, and the QCode setup.
Crypto payment guide
Nine coins, 56 networks, zero platform fee for AI coding API credits.
The figures and quotes on this page were checked on 2026-11-09 against OpenCode's official pages, which take precedence; changes may happen without notice. Estimated request counts are OpenCode's estimates from typical usage, not a measure of your own usage. Not affiliated with OpenCode or DeepSeek. Model availability on QCode is whatever the /models page shows.