Market2026-06-30

Reports say DeepSeek V4 formal release is planned for mid-July, with peak-hour API prices doubling during Beijing work blocks

Checked on June 30, 2026: Chinese financial and technology outlets report that DeepSeek upgrade emails point to a mid-July V4 formal release and a peak-valley API pricing policy. Current public DeepSeek docs still show the flat V4 Pro and Flash price table, so this item treats the peak-hour policy as a reported pricing update rather than a storefront or inventory change.

Reported DeepSeek V4 pricing update

Chinese media coverage on June 29 says DeepSeek is preparing a formal DeepSeek V4 release for mid-July and will introduce peak-valley API billing after that release. This is a pricing and capacity-management story, not a new stocked Coding Plan listing for this site.

What the current reports say

  • Formal V4 timing: reports citing DeepSeek upgrade reminders say the formal DeepSeek V4 release is planned for mid-July 2026.
  • Peak windows: the reported high-demand windows are Beijing time 9:00-12:00 and 14:00-18:00.
  • Peak multiplier: reports say V4 Pro and V4 Flash keep their usual prices outside peak windows, while peak-hour API prices become 2x the usual rates.
  • V4 Pro peak examples: reported peak-hour V4 Pro pricing is 0.05 yuan per 1M cached-input tokens, 6 yuan per 1M uncached-input tokens, and 12 yuan per 1M output tokens.
  • V4 Flash peak examples: reported peak-hour V4 Flash pricing is 0.04 yuan per 1M cached-input tokens, 2 yuan per 1M uncached-input tokens, and 4 yuan per 1M output tokens.
  • Reason given in reports: the change is framed as a way to allocate resources more reasonably and improve service stability during high-load work hours.

Official docs baseline checked today

DeepSeek's public Chinese Models & Pricing page still showed the current flat table when checked for this update: V4 Flash at 0.02 yuan cached input, 1 yuan uncached input, and 2 yuan output per 1M tokens; V4 Pro at 0.025 yuan cached input, 3 yuan uncached input, and 6 yuan output per 1M tokens. The same official page lists 1M context, 384K maximum output, and current concurrency limits of 2500 for Flash and 500 for Pro.

The official page also says prices may change and asks users to check that page for the latest pricing. That matters editorially: until DeepSeek publishes a public peak-valley table on its own docs, this site should write the 2x peak policy as a reported change from upgrade emails and user confirmations, while keeping the current docs table as the public baseline.

Why this matters for API buyers

The reported policy does not make every DeepSeek workload twice as expensive. It shifts the budgeting question from only model choice to time-of-day routing:

  1. Heavy batch jobs, indexing, offline evaluation, and non-urgent agent loops can be scheduled outside Beijing work peaks when possible.
  2. Interactive IDE, coding-agent, customer-support, and office-hour traffic may see higher effective costs if it lands inside the two peak windows.
  3. V4 Flash remains the lower-cost route for high-volume routine traffic, but peak-hour Flash is still reported to be priced at 2x the current flat rate.
  4. V4 Pro remains the route to reserve for harder reasoning, long-context planning, and tasks where the quality difference justifies the higher bill.

Storefront boundary

This news item does not change SUBSCRIPTION_PLANS, stock, payment configuration, or any purchasable /pricing card. The site's Coding Plans are one-off inventory products. API token pricing news can affect buyer expectations, but it does not by itself create a new plan or prove new resellable key inventory.

Rejected or watch-list candidates

  • Immediate pricing-card changes: rejected. This is API pricing news, not confirmed local inventory or storefront stock.
  • Calling the peak table fully public in official docs: rejected. The public docs checked today still show the current flat table and a general price-change notice.
  • Treating the policy as only a rumor: rejected. Multiple mainstream Chinese outlets report the same mid-July and peak-window details, and one report cites multiple API users who received DeepSeek upgrade emails.
  • Assuming global local-time peaks: rejected. The reported schedule is Beijing time.

Editorial takeaway

The practical update is simple: DeepSeek V4's formal release is now being reported for mid-July 2026, and the likely pricing story is not a blanket list-price reset. It is a peak-valley mechanism where Beijing work-hour API calls cost more, while the existing public price table remains the baseline for non-peak planning until DeepSeek's docs expose the new table directly.

Sources checked

View original source