查看中文版 →

Deutsche Version ansehen →

Открыть русскую версию →

Ver versión en español →

Log inChoose a plan
Blog · Costs and limits

DeepSeek's Web Code Points to a 'Coding Plan,' but Nothing Is Announced

Front-end commits on chat.deepseek.com add five-minute quota polling, exhausted-limit banners, and a 'Coding Plan' reference—yet DeepSeek has published no pricing, tiers, or release date.

On November 14, IT Home reported that developers inspecting the front-end code of chat.deepseek.com uncovered a full usage-quota management system pointing to a package named Coding Plan. So far, DeepSeek has not announced any subscription pricing, tiers, or launch date.

The short version: Front-end code on DeepSeek's web chat does contain quota logic and a Coding Plan label, but there is no checkout page, no price, and no release schedule. Treat it as unannounced.

What appeared in the front-end code

According to developers cited by IT Home, recent front-end commits on chat.deepseek.com include user-facing quota messaging—such as telling users their usage quota has reached its limit and to try again once it recovers—alongside state checks.

Specifically, the code covers four behaviors:

  • Five-minute quota polling: While the tab is visible, the client syncs each model's remaining quota, user mute status, and server-supplied copy every five minutes.
  • Exhausted-limit banners and home-page takeover: When a quota runs out, a banner appears above the input box and the home-page welcome area switches to quota notices sent by the server.
  • Error branches and array checks: The commits add an error-handling branch for exhausted quotas and check an is_exhausted flag inside a quotas array when continuing a shared conversation.
  • Recovery and limit warnings: The client handles real-time remaining-quota updates, immediate depletion alerts, recovery notices, and server-side usage-limit warnings.

What is missing from the code

Client-side code alone does not tell you what product—if any—will ship. Reports note that the commits contain no payment screen, no price, and no tier structure, and they do not specify whether the quotas apply to free users, paying users, or both.

Code shipped in a web front end can also belong to internal tests that never launch in that form.

As for community speculation that DeepSeek will debut its first consumer paid plan alongside DeepSeek V4.1 Pro, IT Home explicitly flags that as user guesswork. DeepSeek has not commented on the quota code.

Why developers expect a Coding Plan

Right now, the only way DeepSeek charges individuals directly is its open platform API, which bills per token used rather than selling a flat monthly web membership.

Screenshot of the official DeepSeek Models & Pricing documentation page listing DeepSeek-V4.1-Flash-1017 and DeepSeek-V4-Pro-0813, with Agent Integrations for DeepSeek Harness, Claude Code, Codex, and OpenCode in the left sidebar
DeepSeek's official Models & Pricing page: individual developers still pay per token via the API, even though the sidebar already lists dedicated agent integrations. Source: DeepSeek API Docs

Elsewhere in China's model ecosystem, dedicated coding subscriptions are already standard. Zhipu AI sells its GLM Coding Plan, and Alibaba Cloud and Baidu AI Cloud both offer developer coding plans. Seeing a Coding Plan label inside DeepSeek's codebase naturally led developers to group it with those offerings.

Recent org and usage numbers have fueled the speculation further. In October 2026, DeepSeek merged its Harness team with its Web, App, and Platform teams into a single product department led by former Harness head Cui Tianyi. Around the same time, industry data put DeepSeek at roughly 130 million monthly active users and over $1 billion in annualized revenue—almost all of it from the API—with no direct consumer billing yet and the only year-over-year MAU dip among the top three.

How to run DeepSeek for coding today

While an official Coding Plan remains unconfirmed, we think the best setup depends on which models your workflow actually needs:

  • DeepSeek only: Going straight to DeepSeek's official API and paying per token is usually the cheapest option—you pay for what you send and never waste an unused monthly bundle.
  • Predictable monthly billing: Cloud providers such as Volcengine Ark and Alibaba Cloud Bailian already bundle DeepSeek into monthly coding plans if you prefer a fixed cap over metered token bills.
  • Mixing DeepSeek with Claude, GPT, or Gemini: CodingPass Go is $10 a month for a $60 shared pool (with up to $40 each for open-weight models like DeepSeek V4.1 Flash), metered at official API prices. If your workload runs almost entirely on open-weight models and rarely touches Claude, a dedicated open-model plan gives you a larger single-model allowance; if you switch across frontier and open models, CodingPass keeps them under one bill.
Boarding

Need DeepSeek and Claude in one plan?

CodingPass Go starts at a month and includes DeepSeek V4.1 Flash alongside Claude, GPT, and Gemini, metered at official API prices.

  • Request ID on every response
  • Full refund if you've used 5% or less
  • No surprise charges
Choose GoBrowse models

$10/mo until you cancel.