DeepSeek V4 Flash
A model by DeepSeek · released 2026-04 · 19 checked facts from 3 sources
Timeline
- DeepSeek released DeepSeek V4 Flash, Apr 2026
- DeepSeek V4 Flash is available on Responses API
- DeepSeek deprecates DeepSeek V4 Flash
Prices
Per price, each value with the dates it was seen. A value seen on one date only is one copy of the page.
No price has changed over the record.
| Price for | Values over time |
|---|---|
| per 1M cached input tokens | $0.0028 |
| per 1M input tokens | $0.14 |
| per 1M output tokens | $0.28 |
What it does
- DeepSeek V4 Flash: 1M context window
- DeepSeek V4 Flash: total parameters 284B
- DeepSeek V4 Flash: active parameters 13B
- DeepSeek V4 Flash: performs on par with V4-Pro on simple Agent tasks
- DeepSeek V4 Flash: max output tokens 384K
- DeepSeek V4 Flash: supports both non-thinking and thinking (default) modes
- DeepSeek V4 Flash: Json Output
- DeepSeek V4 Flash: Tool Calls
- DeepSeek V4 Flash: Chat Prefix Completion (Beta)
- DeepSeek V4 Flash: FIM Completion (Beta), non-thinking mode only
- DeepSeek V4 Flash: concurrency limit 2500
- DeepSeek V4 Flash: flexible reasoning effort (low, high, max)
- DeepSeek V4 Flash: parameters 284-billion
In the news
- unemploymentlabs.com: Chinese frontier AI models carry the fingerprint of export-controlled GPUs
- dwarfstar.sh: From the creator of Redis; run LLM locally with ds4
Built 2026-11-21 06:50 UTC from the news
knowledge graph by code. Every line is a fact whose quote was checked against the source's own words (hover a source);
it says who states it, not that it is true. Archived pages are the lab's pages as the Wayback Machine holds them.