DeepSeek V4 Pro

deepseek/deepseek-v4-pro

text in · text out · 1,000,000 context · 384,000 max output

toolsschemareasoningcacheno-train
Effective cost per 1k requests

cost = p_in·(input − cached) + p_cache_read·cached + p_out·output + p_req; cached tokens are the prefix a sticky endpoint already holds (part2b §9.3), which is why the cheapest list price is not always the cheapest request.

ProviderRegion · quantContext / max outInput / MOutput / MCache read / writeMarkupPer 1k reqUptime 30 dTTFT p50 / p95ThroughputLast 24 hFidelityDataCapabilitiesStatus
DeepSeekcn · none1,000,000 / 384,000$1.320$3.960$0.044 / —pass-through$3.344———no traffic— 365 dmodel defaultactive
OpenRouterglobal · unknown1,024,000 / 384,000$0.219$0.439$0.019 / —+5.0%$0.458———no traffic— no-train 30 d16 differactive

Quick start

curl https://api.ai.ml/v1/chat/completions \
  -H "Authorization: Bearer $AIML_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek/deepseek-v4-pro","messages":[{"role":"user","content":"Say hello"}]}'
Get a key in the console

Changelog

  • new_model endpoint ep_openrouter_deepseek-v4-pro_global (deepseek/deepseek-v4-pro) added as active; price input 0.219 output 0.439 USD/M 10/5/2026