DeepSeek has officially released the V4.1 Flash model on September 10, 2026 (Beijing Time). In the meanwhile, we plan to postpone the discontinuation of the V4 Pro service to 12:00 Beijing Time on September 14, 2026. At that time, all requests to the Pro model will be routed to V4.1 Flash and billed at Flash's price. After extensive internal and external testing, V4.1 Flash has comprehensively surpassed V4 Pro across all key metrics, including performance, cost, speed, and task completion time. If you encounter any issues during your comparative testing between V4 Pro and V4.1 Flash, please do not hesitate to reach out to us with your feedback.
The price of V4 Pro remains unchanged during the service period. The price of V4.1 Flash will take effect at 4:00 UTC, September 10, 2026.
Peak hours (in UTC): 1:00–4:00 AM and 6:00–10:00 AM, Monday to Friday. (All other hours are off-peak)
If you continue to use our services after the billing adjustment, you will be deemed to have accepted the adjusted billing terms. If you do not agree, you may choose to cancel your service and apply for a refund. Should you have any questions or require further information, please do not hesitate to contact us.
AlexWApp•3h ago
The architecture is probably part of the explanation for the lower cost and faster inference. DeepSeek says V4.1 Flash uses a new Causal Encoder–Decoder design, with 8B active parameters for input processing and 16B for decoding, along with much smaller KV caches.
But I hope it is just not benchmaxxed and genuinely good model
benchmarks: https://media2url.com/m/52a77a33347c48