DeepSeek-V4-Flash pricing
A million-token context window at a fraction of any Western model's price, with cache-hit input billed at a fiftieth of the base rate. DeepSeek has published notice of a significant price rise, so treat these rates as a floor rather than a plan.
List prices checked
By DeepSeek · Balanced tier
Token pricing
Per million tokens, from the official price list.
| Token type | Price per 1M | Notes |
|---|---|---|
| Input | $0.140 | Everything you send, including the system prompt |
| Output | $0.280 | Everything the model generates, reasoning tokens included |
| Cached input | $0.0028 | Repeated prefixes served from cache |
| Blended (3:1) | $0.175 | One number for ranking, at three input tokens per output token |
Source: DeepSeek pricing (August 2026). Published list prices only — committed-use discounts, regional variation, and per-request charges are not included. Last checked against the vendor’s page on 2026-08-07.
Capabilities
| Property | Value |
|---|---|
| Context window | 1M tokens |
| Max output | 384K tokens |
| Inputs accepted | text |
| Cost to fill the context once | $0.14 |
What it costs per month
Token prices applied to three workload shapes. Input-heavy and output-heavy workloads rank models very differently.
In-app chat assistant
A support or help assistant handling a few thousand conversations a month, with short prompts and short answers.
$4.20/mo
- 20M input tokens$2.80
- 5M output tokens$1.40
- If fully cached$1.46
RAG over documents
Retrieval-augmented answering where each request stuffs several retrieved passages into the prompt, so input dominates.
$46/mo
- 300M input tokens$42
- 15M output tokens$4.20
- If fully cached$5.04
Coding agent
An agentic loop that reads files and writes patches, generating heavy output and re-sending large context on every turn.
$84/mo
- 400M input tokens$56
- 100M output tokens$28
- If fully cached$29
DeepSeek-V4-Flash vs other balanced models
Same capability tier, ranked by blended token price.
| Model | Input | Output | Context | Compare |
|---|---|---|---|---|
| GPT-5 mini | $0.250 | $2.00 | 400K | Head to head |
| Gemini 2.5 Flash | $0.300 | $2.50 | 1M | Head to head |
| Grok 3 mini | $0.300 | $0.500 | 131K | Head to head |
| Codestral | $0.300 | $0.900 | 262K | Head to head |
| Claude Haiku 4.5 | $1.00 | $5.00 | 200K | Head to head |
| OpenAI o4-mini | $1.10 | $4.40 | 200K | Head to head |
Questions about pricing
- How much does the DeepSeek-V4-Flash API cost in 2026?
- $0.140 per million input tokens and $0.280 per million output tokens, from DeepSeek's published pricing. The table above converts that into a monthly figure for three common workloads.
- What is the DeepSeek-V4-Flash context window?
- 1M tokens, with up to 384K tokens of output per request. Filling the window costs whatever that many input tokens cost — at $0.140 per million, a single full-context request runs about $0.14.
- Does DeepSeek-V4-Flash support prompt caching?
- Yes. Cached input is billed at $0.0028 per million tokens, roughly 50× cheaper than uncached input. For an agent that resends the same system prompt and file context on every turn, that discount usually matters more than the headline rate does.
- Is DeepSeek-V4-Flash cheaper than the alternatives?
- On a blended 3:1 input-to-output basis it works out at $0.175 per million tokens. The comparison table below ranks it against the other balanced models in the dataset, and the head-to-head pages model both against identical workloads.