DeepSeek V4 Flash
deepseek · deepseek/deepseek-v4-flash
Fast DeepSeek V4 lane for economical reasoning, coding, and long-context work
≈667k Chinese characters of context · Step-by-step reasoning · Calls tools / agents · Structured JSON output · Open weights, self-hostable
Read a 200,000-Chinese-character book and write a 2,000-Chinese-character summary: about $0.05
An order-of-magnitude estimate at 1.5 tokens per Chinese character and the vendor's own rate — not a quote. Full method: the glossary on the AI Models page.
Specs
| Context | 1,000,000(≈667k Chinese characters) |
|---|---|
| Max output | 384,000 |
| Structured output | Yes |
| Open weights | Yes |
| Released | 2026-04-24 |
Source: models.dev snapshot 2026-10-12
Who serves it, at what price
| Provider | In /M | Out /M | Cache read /M |
|---|---|---|---|
| DeepSeekfirst-party | $0.15 | $0.6 | $0.003 |
| above.dev | $0.165 | $0.66 | $0.0033 |
| AgentRouter | — | — | — |
| Alibaba (China) | $0.14 | $0.28 | $0.0028 |
| Alibaba Token Plan | Billed by plan | Billed by plan | $0 |
| Alibaba Token Plan (China) | Billed by plan | Billed by plan | $0 |
| Auriko | $0.14 | $0.28 | $0.0028 |
| Azure | $0.19 | $0.51 | — |
| CrofAI | $0.12 | $0.21 | $0.003 |
| InferX | Billed by plan | Billed by plan | — |
| Kenari | Billed by plan | Billed by plan | — |
| DevPass (LLM Gateway) | $0.065 | $0.116 | $0.012 |
| Modelis | $0.0983 | $0.1966 | — |
| NaN | Billed by plan | Billed by plan | — |
| Neuralwatt | $0.14 | $0.28 | $0.028 |
| Ollama Cloud | $0.22 | $0.66 | $0.007 |
| OpenCode Zen | $0.14 | $0.28 | $0.028 |
| OpenCode Go | $0.15 | $0.6 | $0.003 |
| Opper | $0.25 | $0.66 | — |
| Pendra | Billed by plan | Billed by plan | — |
| Requesty | $0.28 | $0.56 | $0.07 |
| routing.run | $0.112 | $0.224 | — |
| SenseNova (China) | Billed by plan | Billed by plan | $0 |
| UnoRouter | $0.0625 | $0.125 | — |
| Venice AI | $0.138 | $0.275 | $0.028 |
| Vivgrid | $0.15 | $0.3 | $0.03 |
| Volcengine Ark Coding Plan | Billed by plan | Billed by plan | $0 |
Rows tagged “first-party” are the model vendor's own pricing; untagged rows are resale channels and may differ.
When to choose it
- 量大、单次任务简单(分类、抽取、改写),想把单价压到最低
- 一方官方价 $0.14/$0.28 每百万 token,同时给到 100 万上下文 —— 这个组合在精选清单里没有第二个
Caveats
- 不接受图片输入,纯文本模型
Basis: first-party pricing page · reviewed 2026-08-09
Leaderboard
Not yet on any leaderboard we track. That does not mean the model performs poorly — it means we have not taken this snapshot for it yet, and we do not publish a score we have not measured.