DeepSeek/deepseek-v4-flash/Released Jul 2026

DeepSeek V4 Flash

Efficient MoE for agents, 1M context

Text

About DeepSeek V4 Flash

DeepSeek V4 Flash is the efficiency variant of the V4 family. The DeepSeek-V4-Flash-0731 build (July 31, 2026) keeps the same 284B total / 13B active MoE architecture as the April preview, with re-post-training that substantially improves agent, coding, and tool-use workloads. Supports a 1M-token context window, thinking and non-thinking modes, structured outputs, and tool use. Pairs naturally with V4 Pro: route the hard prompts to Pro, everyday and agent traffic to Flash.

MetricValue
Parameter Count284 billion (13 billion active)
Mixture of ExpertsYes
Active Parameter Count13 billion
Context Length1,048,576 tokens
Max Output384,000 tokens
MultilingualYes
Tool UseYes
Structured OutputsYes

Ready to build with DeepSeek V4 Flash?

Try DeepSeek V4 Flash in the Workbench to prompt it, compare outputs, and iterate on prompts without writing any code. When you're ready to ship, call the same model from our API and build your own apps on top of it.

curl -sSf -X POST https://hub.oxen.ai/api/ai/chat/completions \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $OXEN_API_KEY" \
    -d '{
  "model": "deepseek-v4-flash",
  "messages": [
    {
      "role": "user",
      "content": "Try sending a message."
    }
  ]
}'

API endpoint

See the API reference for request and response formats.

POSThttps://hub.oxen.ai/api/ai/chat/completions