Handles difficult analysis, planning, and multi-step problem solving.
Back to models
Deep reasoning Large-context engineering Agent integration
DeepSeek
DeepSeek V4 Pro
Deep reasoning for demanding workloads47% offInputTextOutputText
API formatOpenAI
A high-capability reasoning model for complex coding, research, and long-running agents.
deepseek-v4-proInput price$0.348$0.66per million tokens
Output price$0.696$1.98per million tokens
Cache read$0.00293$0.022per million tokens
Context window1M tokensMaximum context per request
Max output384K
BillingPer token
HealthOperationalSuccess 100% · latency 5.0s · throughput 42.9
Model details
Capabilities, benchmarks, and typical use
DeepSeek V4 Pro is the high-capability member of the DeepSeek V4 family. The local DeepSeek adapter exposes it as a text-in/text-out model with reasoning, tool calling, structured output, and a 1M-token context window. The original prices use the active local new-api pricing configuration, including the configured cache-read price.
Processes repositories, specifications, and long-running task histories in one context.
Combines function calling, structured output, and reliable tool workflows.