What does the V4 hybrid attention architecture change for inference?

Deepseek V4 Flash combines Compressed Sparse Attention (CSA) with Heavily Compressed Attention (HCA), and uses ManifoldConstrained Hyper-Connections (mHC) in place of standard residual connections. The combination targets efficient inference at long context, including the full 1.0M tokens window.

When should I pick Deepseek V4 Flash over DeepSeek V4 Pro?

Pick Deepseek V4 Flash for instruction following, classification, and short-form question answering where latency and per-token cost matter most. Use DeepSeek V4 Pro for complex reasoning, multi-step problem solving, and agentic tasks.

What is the context window and max output for Deepseek V4 Flash?

The context window is 1.0M tokens and the maximum output is 1.0M tokens.

What does implicit caching do for pricing?

Implicit caching detects repeated input prefixes (typically long system prompts) and charges the cached input rate of $0.007 per token instead of the standard $0.08 input rate. No explicit cache-control header is required.

Does Deepseek V4 Flash support tool calls?

Yes. Deepseek V4 Flash is tagged for tool use and reasoning, so function calling works through the AI SDK as well as Chat Completions, Responses, and Messages API formats.

Does Deepseek V4 Flash support zero data retention?

Yes, Zero Data Retention is available for this model. Zero Data Retention is offered on a per-provider basis. See https://vercel.com/docs/ai-gateway/capabilities/zdr for details.

DeepSeek

Deepseek V4 Flash

Deepseek V4 Flash is DeepSeek's April 23, 2026 efficiency-tier model in the V4 series. It pairs a hybrid attention architecture with a context window of 1.0M tokens and supports reasoning, tool use, and implicit caching.

View API reference

Input and output price: Prices from: Input $0.08, Output $0.18, Per 1M tokens
24h uptime: Loading AI Gateway uptime

Get API key Read the docs

import { streamText } from 'ai'

const result = streamText({
  model: 'deepseek/deepseek-v4-flash',
  prompt: 'Why is the sky blue?'
})

Read docs

Overview Playground Providers Uptime Status Throughput Latency API Similar About FAQ

Copy link to headingPlayground

Try out Deepseek V4 Flash by DeepSeek. Usage is billed to your team at API rates. Free users (those who haven't made a payment) get $5 of credits every 30 days.

Deepseek V4 Flash

Copy link to headingProviders

Route requests across multiple providers. Copy a provider slug to set your preference. Visit the docs for more info. Using a provider means you agree to their terms, listed under Legal.

Provider

Context	Max Output	Latency	Throughput	Input	Output	Cache	Web Search	Capabilities	ZDR	No Training	Regional Inference	Free Tier	Release Date

DigitalOcean

Legal:Terms•Privacy

262K

52K

0.5 s

201 tps

$0.11/M

$0.22/M

Read$0.03/M

—

04/23/2026

DeepSeek

Legal:Terms•Privacy

384K

1.4 s

142 tps

$0.22/M+1 more

$0.66/M+1 more

Read$0.007/M

—

04/23/2026

Novita AI

Legal:Terms•Privacy

393K

1.2 s

124 tps

$0.44/M

$1.32/M

Read$0.03/M

—

04/23/2026

DeepInfra

Legal:Terms•Privacy

1.2 s

26 tps

$0.08/M

$0.18/M

Read$0.02/M

—

04/23/2026

Fireworks

Legal:Terms•Privacy

384K

3.7 s

95 tps

$0.22/M+1 more

$0.66/M+1 more

Read$0.007/M

—

04/23/2026

Azure

Legal:Terms•Privacy

128K

3.3 s

109 tps

$0.19/M

$0.51/M

Read$0.03/M

—

04/23/2026

Alibaba Cloud

Legal:Terms•Privacy

384K

3.3 s

107 tps

$0.22/M+1 more

$0.66/M+1 more

Read$0.02/M

—

04/23/2026

Baseten

Legal:Terms•Privacy

384K

4.9 s

85 tps

$0.13/M

$0.26/M

Read$0.03/M

—

04/23/2026

Wafer

Legal:Terms•Privacy

0.8 s

223 tps

$0.28/M

$0.56/M

Read$0.07/M

—

04/23/2026

GMICloud

20% off

Legal:Terms•Privacy

384K

1.9 s

$0.35/M

$1.06/M

Read$0.01/M

—

04/23/2026

Getting started

Call Deepseek V4 Flash through AI Gateway with the AI SDK generateText and streamText functions, or through the OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages APIs by changing the base URL. AI Gateway authenticates the request and routes it to an available provider.

Install the AI SDK (pnpm add ai dotenv), create an API key from the API Keys page, and set it as AI_GATEWAY_API_KEY in your environment. Full setup is covered in the text generation quickstart.

index.ts

import { generateText } from 'ai';
import 'dotenv/config';

async function main() {
  const result = await generateText({
    model: 'deepseek/deepseek-v4-flash',
    prompt: 'Why is the sky blue?',
  });

  console.log(result.text);
}

main().catch(console.error);

Top-level parameters

The same Deepseek V4 Flash request in each API format AI Gateway supports.

top-level-params.ts

import { generateText } from 'ai';
import 'dotenv/config';

async function main() {
  const result = await generateText({
    model: 'deepseek/deepseek-v4-flash',
    system: 'You are a concise technical assistant.',
    prompt: 'Summarize the tradeoffs between static generation and SSR.',
    maxOutputTokens: 1024,
  });

  console.log(result.text);
}

main().catch(console.error);

Standard parameters like prompt, messages, temperature, and tools work as documented in the AI SDK docs. These are the parameters with model-specific behavior.

Parameter	Type	Required	Description
`model`	`string`	Yes	Model ID in the form `creator/model`, e.g. `deepseek/deepseek-v4-flash`. AI Gateway routes the request to an available provider.
`maxOutputTokens`	`number`	No	Hard cap on generated tokens. Deepseek V4 Flash supports up to `1,048,576` output tokens. Reasoning tokens count toward this limit.
`reasoning`	`'provider-default' \| 'none' \| 'minimal' \| 'low' \| 'medium' \| 'high' \| 'xhigh'`	No	Provider-agnostic reasoning effort, available in AI SDK 7 or later. Maps to the provider’s native reasoning configuration; reasoning settings under `providerOptions` take precedence when both are set. See the Reasoning section below.
`providerOptions`	`Record<string, JSONValue>`	No	AI Gateway routing options under `gateway`, plus any provider-native options under the provider’s own namespace — see the table below.

Input limits

Input	Formats	Sources	Max count	Max size	Limits
Text	—	—	—	—	Prompt and response share the 1M-token context window

Provider options

Set AI Gateway routing options under providerOptions.gateway. For provider-specific options, pass them under the provider’s namespace as documented by the AI SDK.

Learn more in the AI SDK deepseek provider docs.

provider-options.ts

import { generateText } from 'ai';
import 'dotenv/config';

async function main() {
  const result = await generateText({
    model: 'deepseek/deepseek-v4-flash',
    prompt: 'Why is the sky blue?',
    providerOptions: {
      gateway: {
        only: ['digitalocean', 'deepseek'],
      },
    },
  });

  console.log(result.text);
}

main().catch(console.error);

These AI Gateway routing options apply to every model. Provider-specific options pass through under the provider’s own namespace (for example providerOptions.anthropic) exactly as documented by the AI SDK.

Parameter	Type	Required	Description
`providerOptions.gateway.only`	`string[]`	No	Restrict routing to these provider slugs. Requests fail over only within the listed providers.
`providerOptions.gateway.order`	`string[]`	No	Preferred provider order. Listed providers are tried first; unlisted providers remain available as fallbacks.
`providerOptions.gateway.sort`	`'cost' \| 'ttft' \| 'tps'`	No	Rank candidate providers by price, time to first token, or tokens per second instead of the default routing order.
`providerOptions.gateway.zeroDataRetention`	`boolean`	No	Route only to providers with a zero-data-retention policy for this model.

Routing across providers

AI Gateway serves the same model through multiple providers and fails over automatically. order expresses a preference while keeping every provider eligible; only is a hard allowlist — if none of the listed providers are available the request fails instead of falling back.

Options under a provider's own namespace (for example providerOptions.anthropic) are forwarded to that provider with the request. Providers ignore option namespaces that don't apply to them, so it is safe to set provider options alongside gateway routing options.

Reasoning

AI Gateway bridges reasoning across every API format. The AI SDK exposes a provider-agnostic top-level reasoning level (none, minimal, low, medium, high, or xhigh); the Chat Completions and Responses formats take the same effort under reasoning.effort; and the Anthropic Messages format uses a native thinking token budget. Whichever you send, the gateway maps it to the target model’s native configuration, converting between effort levels and token budgets as needed. Reasoning-related settings under providerOptions take full precedence over the top-level reasoning value and are never merged. Reasoning tokens typically count toward your output-token usage, though how they’re reported and billed varies by provider.

Learn more in the AI Gateway reasoning guide.

reasoning.ts

import { generateText } from 'ai';
import 'dotenv/config';

async function main() {
  const result = await generateText({
    model: 'deepseek/deepseek-v4-flash',
    prompt: 'Explain the Monty Hall problem step by step.',
    reasoning: 'high',
  });

  console.log(result.text);
}

main().catch(console.error);

Tool calling

Expose tools the model can call. Define each tool’s inputs with a Zod schema.

tool-calling.ts

import { generateText, tool } from 'ai';
import { z } from 'zod';
import 'dotenv/config';

async function main() {
  const result = await generateText({
    model: 'deepseek/deepseek-v4-flash',
    prompt: 'What is the weather in San Francisco?',
    tools: {
      getWeather: tool({
        description: 'Get the current weather for a location',
        inputSchema: z.object({ location: z.string() }),
        execute: async ({ location }) => ({ location, temperatureC: 18 }),
      }),
    },
  });

  console.log(result.text);
}

main().catch(console.error);

Copy link to headingMore models by DeepSeek

Model

Context	Latency	Throughput	Input	Output	Cache	Web Search	Capabilities	Providers	ZDR	No Training	Free Tier	Release Date

deepseek/deepseek-v4.1-flash-beta

1.3 s

331 tps

$0.22/M

$0.66/M

Read$0.007/M

—

09/08/2026

deepseek/deepseek-v4-flash-vision-exp

0.4 s

172 tps

$0.22/M

$0.65/M

Read$0.007/M

—

08/21/2026

deepseek/deepseek-v4-pro-0813

0.3 s

107 tps

$0.60/M

$1.90/M

Read$0.02/M

—

08/12/2026

deepseek/deepseek-v4-flash-0731

0.5 s

244 tps

$0.07/M

$0.15/M

Read$0.007/M

—

+16

07/31/2026

deepseek/deepseek-v4-pro

0.6 s

139 tps

$0.66/M

$1.98/M

Read$0.02/M

—

04/23/2026

deepseek/deepseek-v3.2

164K

0.4 s

124 tps

$0.28/M

$0.42/M

Read$0.03/M

—

12/01/2025

Copy link to headingAbout Deepseek V4 Flash

Deepseek V4 Flash was released April 23, 2026 as part of DeepSeek's V4 generation. The V4 series introduces a hybrid attention architecture that combines Compressed Sparse Attention (CSA) with Heavily Compressed Attention (HCA), along with ManifoldConstrained Hyper-Connections (mHC) that refine standard residual connections. The combination targets efficient long-context inference at the 1.0M tokens window.

Deepseek V4 Flash positions as the efficiency tier of the V4 lineup. It handles instruction following, classification, short-form Q&A, and other tasks where latency and per-token cost matter more than maximum reasoning depth. Maximum output is 1.0M tokens, the same budget as DeepSeek V4 Pro, so single-call response length is not the differentiator. The split between Flash and Pro is about capability depth and cost.

Deepseek V4 Flash supports tool use and reasoning, and the model is tagged for implicit caching. Implicit caching reduces input-token charges for repeated prefixes without requiring explicit cache-control headers in the request. Access is through AI Gateway with an AI Gateway API key or OIDC token, so you don't need a separate DeepSeek platform account.

Copy link to headingWhat To Consider When Choosing a Provider

Configuration: Deepseek V4 Flash is tuned for speed and cost on shorter tasks. If your workload involves multi-step reasoning, complex agentic flows, or long synthesis chains, DeepSeek V4 Pro is the better fit within the same generation.
Zero Data Retention: Zero Data Retention is available for this model. It is offered on a per-provider and model basis. See the documentation for details.
Authentication: AI Gateway authenticates requests using an API key or OIDC token. You do not need to manage provider credentials directly.

Copy link to headingWhen to Use Deepseek V4 Flash

Best for

High-volume classification: Routing and short-answer pipelines where $0.08 input and $0.18 output keep unit economics tight
Short-form instruction following: Summarization of short inputs, structured extraction, and rewriting tasks without multi-step planning
Front-line agent steps: Intent detection and parameter parsing before handing off to a deeper-reasoning model
Implicit caching workloads: Long, repeated system prompts across many calls benefit from cached input pricing

Consider alternatives when

Complex agent orchestration: Use DeepSeek V4 Pro within the same generation for multi-step reasoning and tool planning
Earlier-generation pricing: DeepSeek V3 family models may be lower cost when the 1.0M tokens window or V4 capabilities aren't required
Dedicated deep reasoning: DeepSeek-R1 remains the open-weights reasoning specialist for extended chain-of-thought workloads

Copy link to headingConclusion

Deepseek V4 Flash is the efficiency tier of the V4 generation, suited to high-volume short-form tasks where cost and latency dominate. For deeper reasoning and agentic workflows within the same generation, step up to DeepSeek V4 Pro.

Copy link to headingFrequently Asked Questions

What does the V4 hybrid attention architecture change for inference?
Deepseek V4 Flash combines Compressed Sparse Attention (CSA) with Heavily Compressed Attention (HCA), and uses ManifoldConstrained Hyper-Connections (mHC) in place of standard residual connections. The combination targets efficient inference at long context, including the full 1.0M tokens window.
When should I pick Deepseek V4 Flash over DeepSeek V4 Pro?
Pick Deepseek V4 Flash for instruction following, classification, and short-form question answering where latency and per-token cost matter most. Use DeepSeek V4 Pro for complex reasoning, multi-step problem solving, and agentic tasks.
What is the context window and max output for Deepseek V4 Flash?
The context window is 1.0M tokens and the maximum output is 1.0M tokens.
What does implicit caching do for pricing?
Implicit caching detects repeated input prefixes (typically long system prompts) and charges the cached input rate of $0.007 per token instead of the standard $0.08 input rate. No explicit cache-control header is required.
Does Deepseek V4 Flash support tool calls?
Yes. Deepseek V4 Flash is tagged for tool use and reasoning, so function calling works through the AI SDK as well as Chat Completions, Responses, and Messages API formats.
Does Deepseek V4 Flash support zero data retention?
Yes, Zero Data Retention is available for this model. Zero Data Retention is offered on a per-provider basis. See https://vercel.com/docs/ai-gateway/capabilities/zdr for details.

Your use is subject to DeepSeek's Terms & Privacy Policies.

Agent Stack

Core Platform

Tools

Learn

Build

Explore

Deepseek V4 Flash

Copy link to headingPlayground

Copy link to headingProviders

Copy link to headingUptime24 hours

Copy link to headingThroughput24 hours

Copy link to headingLatency24 hours

Getting started

Top-level parameters

Input limits

Provider options

Routing across providers

Reasoning

Tool calling

Copy link to headingMore models by DeepSeek

Copy link to headingAbout Deepseek V4 Flash

Copy link to headingWhat To Consider When Choosing a Provider

Copy link to headingWhen to Use Deepseek V4 Flash

Best for

Consider alternatives when

Copy link to headingConclusion

Copy link to headingFrequently Asked Questions