DeepSeek said on 2026-09-10 that at 04:00 UTC on September 14, calls to deepseek-v4-pro will be served by V4.1-Flash and billed on Flash rates—same request name, different model and cost. Teams should re-eval tool use, schemas, latency, and spend before that cutover, not after silent drift shows up in production.
Per the DeepSeek announcement, the swap is temporary until V4.1-Pro launches. Older deepseek-v4-flash and deepseek-v4-flash-vision-exp names are also being retired, but still resolve to V4.1-Flash for now so existing integrations do not break immediately.
DeepSeek says V4.1-Flash beats V4-Pro on some internal tests, but those results are vendor-reported. Pricing needs care too: OrcaRouter’s analysis argues the Flash card itself may be unchanged, with the clearest savings when former Pro traffic lands on Flash rates.