Skip to content

status.openkey.ai is live — independent, worldwide uptime proof See it now ->

openkey

DeepSeek

11 models built · 2 served

DeepSeek is a Chinese AI lab (Hangzhou DeepSeek Artificial Intelligence Co., Ltd.), founded in 2023, known for the V3 and R1 open-weight model families that hit frontier-level performance at unusually low training cost. On OpenKey it offers 11 models spanning three lines: the V3/V3.1/V3.2 chat series for general use, the R1 series for reasoning-heavy tasks, and a distilled R1-on-Llama-70B variant for cheaper inference. The newest release, DeepSeek V4 Flash, pushes context to over 1M tokens. None of DeepSeek's models are free on OpenKey, but pricing across the lineup stays well below most frontier competitors, which is the core of their strategy: near-frontier capability at commodity prices.

Lineup

For general chat and coding, DeepSeek V3.2 ($0.2288/$0.3432 per 1M in/out, provider price) is the current mainline model with a 131K context window. If you need deep reasoning (math, multi-step logic), use R1 0528 ($0.50/$2.15 per 1M) — it's the most capable reasoning model in the lineup, though pricier on output. For the cheapest option, DeepSeek V4 Flash runs $0.09/$0.18 per 1M with a 1,048,576-token context window, making it the best pick for large-document or long-context jobs on a budget. Skip the older R1 Distill Llama 70B ($0.80/$0.80) unless you specifically need a distilled, self-hostable-style checkpoint.

  • DeepSeek V4 Flash ctx 1.0M · in $0.093 · out $0.185
  • DeepSeek V4 Pro ctx 1.0M · in $0.448 · out $0.896
  • DeepSeek V3.2 ctx 131K · in $0.236 · out $0.353
  • DeepSeek V3.2 Exp ctx 164K · in $0.278 · out $0.422
  • DeepSeek V3.1 Terminus ctx 164K · in $0.278 · out $0.979
  • DeepSeek V3.1 ctx 164K · in $0.216 · out $0.814
  • R1 0528 ctx 164K · in $0.515 · out $2.21
  • DeepSeek V3 0324 ctx 164K · in $0.247 · out $0.927
  • R1 Distill Llama 70B ctx 128K · in $0.824 · out $0.824
  • R1 ctx 164K · in $0.721 · out $2.58
  • DeepSeek V3 ctx 131K · in $0.206 · out $0.824

Prices per 1M tokens, flat 3% fee included.

Models DeepSeek serves

Beyond its own models, DeepSeek hosts 2 models in the OpenKey catalog. Prices are DeepSeek’s own list price per 1M input tokens, before our fee.

ModelBuilt byContextInput /M
DeepSeek V4 FlashDeepSeek1.0M$0.140
DeepSeek V4 ProDeepSeek1.0M$0.435

From the OpenRouter catalog, Jul 28, 2026. All providers

Status page ↗Privacy policy ↗Terms of service ↗

Questions

What does DeepSeek's API cost through OpenKey?
Prices vary by model. Cheapest is DeepSeek V4 Flash at $0.09 input / $0.18 output per 1M tokens; most expensive is R1 0528 at $0.50 input / $2.15 output. OpenKey adds a flat 3% fee, so V4 Flash comes to $0.0927 / $0.1854 per 1M ($0.09 × 1.03, $0.18 × 1.03), and R1 0528 comes to $0.515 / $2.2145 per 1M.
Which DeepSeek model should I use for reasoning tasks?
R1 0528 is DeepSeek's dedicated reasoning model, priced at $0.50 input / $2.15 output per 1M tokens (provider price), with a 163,840-token context window. It's the successor to the original R1 (released January 2025) and the strongest reasoning option in the current 11-model lineup.
Are any DeepSeek models free on OpenKey?
No. DeepSeek has 0 free models in its 11-model lineup on OpenKey, unlike some other labs on the platform (25 free models exist across all 52 labs combined). Every DeepSeek model, from the cheapest V4 Flash to the priciest R1 0528, carries per-token provider pricing plus OpenKey's flat 3% fee.