Updated 8 Oct 2026: Correct material V4 Pro routing and billing contradiction using current DeepSeek changelog/rate card; original reroute announcement remains online but conflicts.

Key details

  1. September 10 announcement said V4 Pro would reroute from September 14.
  2. Current official API Change Log says V4 Pro continues after September 14 with unchanged billing.
  3. Current official rate card identifies deepseek-v4-pro as DeepSeek-V4-Pro-0813, distinct from deepseek-flash / V4.1 Flash.
  4. Peak Pro cache-miss input/output prices are $1.32/$3.96 per million tokens versus Flash $0.30/$1.20.
  5. Off-peak prices are half peak prices, with published weekday UTC peak windows.
  6. No authenticated API response or billing test was performed; DeepSeek's own pages remain inconsistent.

What builders should take away

  1. Use current API Change Log and Models & Pricing to plan V4 Pro costs, not the older reroute assertion.
  2. Do not assume deepseek-v4-pro now bills at Flash rates.
  3. Verify actual backend metadata and billed tokens before production migration.
  4. Retest workloads when aliases or pricing change; retain peak/off-peak scheduling separately.

What changed

The September 10 V4.1 Flash announcement said all deepseek-v4-pro requests would route to Flash at Flash rates starting September 14. DeepSeek's official API Change Log now explicitly says it decided, in response to user demand, to continue V4 Pro API service after September 14 with unchanged billing. The current Models & Pricing page lists deepseek-v4-pro as DeepSeek-V4-Pro-0813, with distinct Pro prices and a 500 concurrency limit. The older product announcement still says the opposite; this is a documented contradiction, not independently verified live routing.

Why it matters

Developers who budgeted Flash rates or migrated away from V4 Pro on the original announcement may be relying on superseded instructions. The current rate card and changelog point to distinct Pro and Flash service and billing. Teams should verify the actual model and charges on their own account, rather than infer routing from the older announcement.

The current changelog contradicts the original transition

The older September 10 product announcement schedules a V4 Pro-to-Flash reroute from September 14. The official API Change Log for the same release says V4 Pro service continues after September 14 with unchanged billing. Neither page explicitly explains the reversal.

The live rate card retains a distinct Pro model and price

The current Models & Pricing page lists deepseek-v4-pro as DeepSeek-V4-Pro-0813. Its peak cache-miss input/output rates are $1.32/$3.96 per million tokens, versus $0.30/$1.20 for Flash. Off-peak rates are half peak. This is published pricing, not a tested invoice.

Flash compatibility aliases are a separate confirmed change

The same current pricing page explicitly says the retired deepseek-v4-flash and deepseek-v4-flash-vision-exp names route to V4.1 Flash. That does not establish the disputed V4 Pro reroute.

Operational checks should separate backend identity and billing

Pin and inspect model metadata where available, compare invoices with the current rate card, and run task-level regression tests. A stable model string does not always imply stable behavior, but a contradictory marketing page does not prove a reroute occurred.

Timeline

2026-08-13

DeepSeek V4 Pro reaches GA

DeepSeek released the production V4 Pro endpoint.
2026-08-17

Peak/off-peak pricing takes effect

DeepSeek introduced time-dependent API pricing.
2026-09-10

V4.1 Flash announced

DeepSeek launched V4.1 Flash with lower pricing and announced the V4 Pro transition.
2026-09-14

Announced reroute contradicted by current documentation

The original product announcement schedules a Flash reroute, but the official changelog now says Pro service continues unchanged.
2026-10-08

Current API pricing rechecked

The live rate card still lists V4-Pro-0813 as a separate model with distinct rates.

What to watch next

  • DeepSeek reconciling the two official release documents.
  • Any authenticated model metadata and billing changes for deepseek-v4-pro.
  • V4.1 Pro launch timing and whether it replaces V4-Pro-0813.
  • Changes to peak/off-peak prices and concurrency.

Still unclear

  • DeepSeek's announcement and changelog disagree on V4 Pro routing.
  • The live rate card indicates distinct Pro service, but no authenticated request or invoice was inspected.
  • The date when the changelog's continuation decision was made is not specified.

Sources

Direct reading behind this dossier.

3 sources
DeepSeek API Change Log
DeepSeek primary

Current September 10 entry says V4 Pro service continues after September 14 with unchanged billing; conflicts with older release announcement.

Discussion

Discussion is reader-contributed. Comments are not part of the BTN dossier or its editorial evidence.

0 visible comments

Join the discussion

Keep comments useful and relevant. Reader contributions may be moderated and are not BTN editorial evidence.

Sign in to comment