DeepSeek V4.1 Flash Arrives With Lower Prices, and V4 Pro Gets a Reprieve
DeepSeek released DeepSeek-V4.1-Flash on September 10, 2026, and changed its API prices the same day at 04:00 UTC. The announcement also said the larger V4 Pro would be phased out. DeepSeek has since reversed that part.
What V4.1 Flash is
DeepSeek calls V4.1 Flash the smallest model in its new architecture family. According to the company it is a 552B-parameter mixture-of-experts model with a new causal encoder-decoder design: 8B parameters are active for input and 16B for output. It accepts images as well as text. DeepSeek says its KV cache needs a quarter of the HBM memory and an eighth of the SSD storage of the previous generation, which matters because cache-hit charges are a large part of agent costs. The weights are published on Hugging Face, so it is an open-weights model.
In the API the model name is deepseek-flash. V4 Flash and V4-Flash-Vision-Exp are retired; the old names deepseek-v4-flash and deepseek-v4-flash-vision-exp still work, but requests go to V4.1 Flash and are billed at the Flash price.
Prices
DeepSeek keeps time-of-day pricing, with off-peak rates at half the peak rates. Peak hours are 01:00-04:00 and 06:00-10:00 UTC, Monday to Friday. Prices per 1M tokens on DeepSeek’s pricing page, checked September 16, 2026:
| Model | Period | Input (cache miss) | Input (cache hit) | Output |
|---|---|---|---|---|
| V4.1 Flash | Off-peak | $0.15 | $0.003 | $0.60 |
| V4.1 Flash | Peak | $0.30 | $0.006 | $1.20 |
| V4 Pro | Off-peak | $0.66 | $0.022 | $1.98 |
| V4 Pro | Peak | $1.32 | $0.044 | $3.96 |
Both models have a 1M-token context window and up to 384K output tokens. The concurrency limit is 2,500 for Flash and 500 for V4 Pro.
V4 Pro: announced phase-out, then reversed
The September 10 announcement said DeepSeek was phasing out V4 Pro, citing tests that put V4.1 Flash ahead of it on performance, cost and speed. From 04:00 UTC on September 14, all deepseek-v4-pro requests were to be served by V4.1 Flash at Flash rates until a V4.1 Pro launched.
The current pricing page says otherwise. “In response to user demand”, DeepSeek will keep providing V4 Pro after September 14, 2026, with billing unchanged, and promises further notice of any change. The model version behind deepseek-v4-pro is listed as DeepSeek-V4-Pro-0813.
Teams that pinned V4 Pro can keep it for now, but DeepSeek has already shown it intends to retire the model once a replacement is ready.
See all DeepSeek rates on our DeepSeek API pricing page and the full spec sheet for DeepSeek V4 Pro.