Skip to content
DeepSeek

Deepseek V4 Flash 0731

Deepseek V4 Flash 0731 is a dated checkpoint of DeepSeek's V4 Flash weights, with notably stronger agentic capabilities than the April preview. Deepseek V4 Flash 0731 scores 82.7 on Terminal-Bench, pairs a hybrid attention architecture with a context window of 1.0M tokens, and routes across Fireworks, DeepSeek, Baseten, DeepInfra, Novita AI, Alibaba Cloud, Wafer, GMICloud on AI Gateway. Your use is subject to DeepSeek's Terms & Privacy Policies.

ReasoningTool UseImplicit Caching
Use with AI GatewayView docs
import { streamText } from 'ai'


const result = streamText({
  model: 'deepseek/deepseek-v4-flash-0731',
  prompt: 'Why is the sky blue?'
})
Read docs
OverviewAPIAboutProvidersThroughputLatencyUptimeStatusSimilarFAQ

Frequently Asked Questions

  • What does the 0731 suffix in deepseek/deepseek-v4-flash-0731 mean?

    It names a dated checkpoint of the weights. deepseek/deepseek-v4-flash-0731 stays on that snapshot, while deepseek/deepseek-v4-flash picks up whatever DeepSeek ships as current.

  • How does Deepseek V4 Flash 0731 compare with the April preview weights?

    Deepseek V4 Flash 0731 has notably stronger agentic capabilities. On Terminal-Bench, Deepseek V4 Flash 0731 scores 82.7, up 25.8 points from 56.9 in the April preview.

  • When should I pin Deepseek V4 Flash 0731 instead of using deepseek/deepseek-v4-flash?

    Pin Deepseek V4 Flash 0731 when you need reproducible behavior, such as stable evaluation baselines or a release you don't want shifting underneath you. Use the undated model ID when you'd rather pick up new weights without a code change.

  • What is the context window and maximum output for Deepseek V4 Flash 0731?

    The context window is 1.0M tokens and the maximum output is 1M tokens per response.

  • What does the V4 hybrid attention architecture change for inference?

    Deepseek V4 Flash 0731 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 across the full context window of 1.0M tokens.

  • Does Deepseek V4 Flash 0731 work in a coding agent?

    Yes. The updated weights target agentic work such as fixing failing tests and opening pull requests. Run vercel ai-gateway coding-agents setup to connect your agents to AI Gateway, then select deepseek/deepseek-v4-flash-0731 in the agent's model configuration.

  • How do I route Deepseek V4 Flash 0731 to a specific provider?

    Pass an order array under providerOptions.gateway with the provider slugs you want tried first. If none of them can serve the request, AI Gateway falls back to the remaining providers.

  • What is the pricing for Deepseek V4 Flash 0731?

    AI Gateway mirrors provider pricing with no markup and adds no platform fee on inference. Rates are listed on this page and shift when providers update their pricing. Individual providers sometimes run promotional rates that expire on a set date and then return to standard rates.

  • Does Deepseek V4 Flash 0731 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.

  • How do I call Deepseek V4 Flash 0731 through AI Gateway?

    Set the model to deepseek/deepseek-v4-flash-0731 in the AI SDK, Chat Completions API, Responses API, Messages API, or other API formats, from TypeScript or Python. Authenticate with an AI Gateway API key or OIDC token, and AI Gateway handles retries and failover across fireworks, deepseek, baseten, deepinfra, novita, alibaba, wafer, gmicloud.