DeepSeek V4-Flash Explained: Retired, and What Replaced It
In short: DeepSeek V4-Flash was a low-cost AI model from DeepSeek, released in public beta on 31 July 2026 and retired on 10 September 2026. Requests that still use its old name, deepseek-v4-flash, are now answered by its successor, DeepSeek V4.1-Flash, and billed at the Flash price - $0.30 per million input tokens and $1.20 per million output at peak, half that off-peak. The current model name is deepseek-flash.
What was DeepSeek V4-Flash?
DeepSeek V4-Flash was the smaller, cheaper sibling of DeepSeek V4-Pro, sold through DeepSeek's API. It entered public beta on 31 July 2026 (DeepSeek's updates page) with a 1M-token context window and up to 384k tokens of output, in thinking or non-thinking mode. DeepSeek is a Chinese AI company; its larger model is covered in DeepSeek V4-Pro explained.
For what a large language model is, see how LLMs work.
Is DeepSeek V4-Flash still available?
No - V4-Flash was retired, and its name now points to DeepSeek V4.1-Flash. DeepSeek's pricing page says to use deepseek-flash as the model name. The legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp are still accepted, but the models behind them have been retired; their requests are served by the DeepSeek-V4.1-Flash-1017 model and billed at the Flash price.
In practice, code that still calls deepseek-v4-flash keeps working, but it is now talking to a different model.
How much does DeepSeek Flash cost now?
DeepSeek bills by time of day: off-peak rates are half the peak rates. Current Flash prices, from DeepSeek's pricing page (read live from our V4.1-Flash model page):
DeepSeek's page lists peak hours as 01:00-04:00 and 06:00-10:00 UTC, Monday to Friday, excluding Chinese public holidays. All other hours are off-peak, including weekends and Chinese public holidays in full. No separate endpoint or setting is needed - the rate depends on when the request arrives. Cache hit input, for repeated text DeepSeek has already processed, is priced much lower again.
What are the Flash model's specs?
| Spec | DeepSeek Flash (served by V4.1-Flash) |
|---|---|
| Developer | DeepSeek |
| Model name | deepseek-flash (legacy deepseek-v4-flash still accepted) |
| Context window | 1M tokens |
| Maximum output | 384K tokens |
| Thinking | Thinking (default) and non-thinking modes |
| Features | JSON output, tool calls, Responses API and Anthropic API formats |
| Vision | Listed as supported for the Flash model on DeepSeek's pricing page |
How does DeepSeek Flash rank?
On LMArena's text leaderboard DeepSeek V4.1-Flash has 1476 Elo (rank 35, max effort). The full ranked table is on our benchmarks page.
Words to know
- Retired model - a model that has been switched off; its name may be redirected to a successor.
- Peak / off-peak pricing - different rates depending on the time a request is sent.
- Cache hit / cache miss - input DeepSeek has already processed (cheaper) versus new input.
- Thinking mode - the model reasons before answering, using extra output tokens.
Related guides
Frequently asked questions
Is DeepSeek V4-Flash discontinued?+
Yes. DeepSeek retired V4-Flash on 10 September 2026. Requests that use the name deepseek-v4-flash are now served by DeepSeek-V4.1-Flash and billed at the Flash price; the current model name is deepseek-flash.
What replaced DeepSeek V4-Flash?+
DeepSeek V4.1-Flash. DeepSeek's pricing page lists the model version behind deepseek-flash as DeepSeek-V4.1-Flash-1017.
How much does DeepSeek Flash cost?+
At peak it costs $0.30 per million input tokens and $1.20 per million output tokens. Off-peak rates are half of the peak rates.
When are DeepSeek's peak hours?+
DeepSeek lists peak hours as 01:00-04:00 and 06:00-10:00 UTC, Monday to Friday, excluding Chinese public holidays. All other hours, including weekends, are billed at the off-peak rate.
What is the context window of DeepSeek Flash?+
The Flash model has a 1M-token context window and can write up to 384K tokens in one reply.
Enjoyed this? Get more in your inbox.
Up to ten of the week's AI news stories, by email once a week.